Technische Universität München TUM School of Life Sciences Lehrstuhl für Systemverfahrenstechnik Following fungal features – Micromorphology and diffusivity of filamentous fungal pellets revealed by three-dimensional imaging and simulation Stefan Schmideder Vollständiger Abdruck der von der TUM School of Life Sciences der Technischen Universität München zur Erlangung des akademischen Grades eines Doktors der Ingenieurwissenschaften (Dr.-Ing.) genehmigten Dissertation. Vorsitzender: Prof. Dr.rer.nat. Philipp Benz Prüfer der Dissertation: 1. Prof. Dr.-Ing. Heiko Briesen 2. Prof. Dr.-Ing. Andreas Kremling 3. Prof. Dr. Peter J. Punt Die Dissertation wurde am 01.04.2021 bei der Technischen Universität München eingereicht und durch die TUM School of Life Sciences am 11.10.2021 angenommen.
102
Embed
Micromorphology and diffusivity of filamentous fungal pellets ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Technische Universität München
TUM School of Life Sciences
Lehrstuhl für Systemverfahrenstechnik
Following fungal features – Micromorphology anddiffusivity of filamentous fungal pellets revealed bythree-dimensional imaging and simulation
Stefan Schmideder
Vollständiger Abdruck der von der TUM School of Life Sciences der Technischen
Universität München zur Erlangung des akademischen Grades eines Doktors der
In 1917, the food chemist James Currie described the filamentous fungus Aspergillus niger as
an efficient producer of citric acid. Only two years later, mass production of citric acid started
and industrial biotechnology was born (Cairns et al., 2018). Further, the discovery of peni-
cillin as a filamentous fungal metabolite in Penicillium chrysogenum (Fleming, 1929) and the
industrialization thereof during World War II was probably the most important breakthrough
of fungal biotechnology and started the antibiotic era. Today, filamentous fungi are still in-
dispensable for the mass production of citric acid (Cairns et al., 2018; Meyer et al., 2020)
and antibiotics (Keller, 2019; Zacchetti et al., 2018), opening multibillion dollar markets and
being a lifesaver for millions of people (Demain, 2014). While large-scale manufacturing
processes have been developed for several products (Meyer et al., 2016), filamentous fungal
biotechnology has emerged as a hope for a sustainable future (Meyer et al., 2020). According
to a thinktank consisting of researchers and companies, filamentous fungal biotechnology can
be key for the transition from a petroleum-based economy into a bio-based circular economy
(Meyer et al., 2020). Thus, this biotech sector will make significant contributions to climate
change mitigation and will meet several United Nations’s sustainable development goals.
Filamentous fungi are favorable hosts for many biotechnological applications (Wösten,
2019) including the production of acids, enzymes, and secondary metabolites (Hoffmeister
and Keller, 2007; Meyer, 2008; Punt et al., 2002). Compared to bacteria, filamentous fungi
benefit from their ability to perform complex post-translational modifications (Wang et al.,
2020), their greatly expanded protein secretion apparatus (Ward, 2012), and their potential
to produce various bioactive molecules (Brakhage, 2013; Keller, 2019; Nielsen et al., 2017).
Further, filamentous fungi are the only microorganisms with the ability to fully degrade lig-
nocellulosic biomass sustainably to a rich and diverse set of useful products (Meyer et al.,
2020).
In industrial biotechnology, filamentous fungi are usually cultivated under submerged con-
ditions. In such processes, fungal morphology affects productivity (Böl et al., 2021; Krull
et al., 2013; Veiter et al., 2018). Generally, filamentous fungi consist of branched tubes called
hyphae and the macromorphology ranges from loose dispersed hyphae to dense hyphal ag-
glomerates called pellets (Papagianni, 2004; Veiter et al., 2018). Both macromorphologies,
dispersed hyphae and pellets, come with some advantages and limitations. Dispersed hyphae
result in a high viscosity of the cultivation broth, and thus, reduce the nutrient supply due to
insufficient mixing (Krull et al., 2010, 2013). Contrary, cultivation broths with pellets as pre-
dominant macromorphology show low viscosities (Cairns et al., 2019b; Krull et al., 2013).
Compared to dispersed hyphae, pellets display enhanced resistance to shear stress (Cairns
et al., 2019b). However, the transport of nutrients and oxygen into pellets is diffusion-limited
1
by the dense structure (Hille et al., 2005, 2009), which alters growth, metabolic activity, and
finally product formation (Cairns et al., 2019b; Krull et al., 2013; Veiter et al., 2018).
The inner structure of pellets, e.g., the spatial distribution of tips and hyphal material,
is known to affect their productivity. However, the micromorphology within whole intact
pellets has not yet been determined. Similarly, there are no correlations between the three-
dimensional (3D) structure and the diffusive mass transport of nutrients, oxygen, and secreted
metabolites within pellets. Both knowledge gaps are caused by the absence of techniques to
measure and analyze the microscopic structure of whole pellets.
In this publication-based dissertation, methods based on X-ray microcomputed tomogra-
phy (µCT) measurements and 3D image analysis were developed to determine the micro-
morphology of whole pellets (Paper I). Further, a method for diffusion computations through
3D images of pellets was developed (Paper II). As counterpart to the measured pellets, a
3D Monte Carlo growth approach was extended to enable the formation of pellets with the
broad morphological range of filamentous microorganisms (Paper III). Diffusion computa-
tions through 66 measured pellets originating from four filamentous fungi and 3125 simu-
lated pellets revealed a universal correlation between the structure and diffusivity of hyphal
networks (Paper III).
Based on the described methods and findings, micromorphologies of whole intact pellets
can be determined and the diffusion of nutrients, oxygen, and secreted metabolites through
dense hyphal networks can be predicted. Applying 3D morphological analysis of pellets,
the outcome of morphological engineering approaches can be investigated in utmost detail.
In addition, existing approaches to model the morphological development of pellets can be
validated and improved through the use of 3D morphological data and the universal diffu-
sion law. Thus, this thesis will contribute to the targeted design of pellet morphologies, i.e.,
morphological engineering, and to enhanced productivities in fungal biotechnology.
2
2 Theoretical background
Cultivated under submerged conditions, the macromorphology of filamentous fungi ranges
from dispersed hyphae to dense hyphal agglomerates called pellets (Krull et al., 2013; Veiter
et al., 2018). Due to limitations in characterizing the structure and mass transport processes
inside pellets, the focus of this dissertation is on pelletized fungi. Because filamentous bacte-
ria and filamentous fungi in sumberged cultures are similar from a morphological perspective
(Böl et al., 2021; Olmos et al., 2013; Zacchetti et al., 2018), the methods to characterize
and model their morphology are often identical. Furthermore, the mass transport of nutrients
and oxygen through both fungal and bacterial pellets is limited by the dense hyphal network
and the inner part of pellets can be starved (Böl et al., 2021; Zacchetti et al., 2018). Due to
these similarities between filamentous fungi and filamentous bacteria, a selection of findings
and methods for filamentous bacteria are also mentioned in this dissertation. In Section 2.1,
the morphology is elaborated. The interplay between morphology and metabolic activity
and resulting morphological engineering approaches for pellets are described in Sections 2.2
and 2.3, respectively.
2.1 Morphology of pellets
This section describes the development of pellets in submerged cultures, morphological prop-
erties, and morphological measurement techniques.
2.1.1 Morphological development: from spore to pellet
Spores play a crucial role for the formation of pellets (Krull et al., 2013; Zhang and Zhang,
2016). Thus, the development of pellets starting from spores is described here.
Filamentous fungal spores are produced under stressful environmental conditions to ensure
the survival of the organism and can be assumed as metabolically inactive cells (Riquelme
et al., 2018). When bioprocesses are inoculated with spores, favorable conditions activate
their metabolism, initiate swelling, and ultimately, result in germ tube formation (Bizukojc
and Ledakowicz, 2006; Paul et al., 1993). Germ tubes represent first hyphal elements and
grow, like all fungal hyphae, based on tip extension. Both germination and the growth of
hyphae are polarized processes, i.e., the formation of new cell composites is directed. Thus,
tubular structures are created (Cairns et al., 2019b; Riquelme et al., 2018). The simultaneous
extension of tips and the development of new tips by branching of hyphae result in filamen-
tous networks. This process leads to an exponential increase of the biomass under unlimited
growth conditions (Krull et al., 2010).
3
Traditionally, it is distinguished between coagulative and non-coagulative pellet formation
types (Metz and Kossen, 1977), which are illustrated in Figure 1. During non-coagulative
pellet formation, spores and germinated spores remain dispersed. For example, Žnidaršic
et al. (1998) showed that Rhizipus stolonifer (synonym R. nigricans) pellets can develop from
single spores. Contrary, coagulative pellets can form from hundreds or thousands of agglom-
erated spores (Fontaine et al., 2010; Metz and Kossen, 1977). A classical representative of
the coagulative type is Aspergillus niger (Cairns et al., 2018). However, different cultivation
conditions cause different morphological behaviors and a final classification of an organism
into coagulative or non-coagulative type is difficult (Veiter et al., 2018; Zhang and Zhang,
2016). For instance, Nielsen et al. (1995) showed that Penicillium chrysogenum exhibit char-
acteristics of both types. In their study, spores remained dispersed whereas branched hy-
phae agglomerated and subsequently developed to pellets. Thus, Veiter et al. (2018) assigned
P. chrysogenum to a third group, the hyphal element agglomerating type.
Figure 1: Development of non-coagulative and coagulative pellets. Spores are marked black,whereas hyphae are grey. Three-dimensional morphologies were simulated with an ownstochastic growth-model that is based on Celler et al. (2012). Scale bar: 100 µm.
2.1.2 Morphological properties
According to Krull et al. (2013), “pellets can be described as stable spherical agglomerates
composed of a branched network of hyphae. Their shape can vary from smooth and spheri-
cal to elongated and hairy” (p. 113). Further, the micromorphology describes the details of
the hyphal network (Krull et al., 2013; Veiter et al., 2018). In this subsection, quantitative
morphological properties of filamentous microorganisms are described together with their
measurement techniques. The advantages and limitations of the mentioned techniques are
elaborated in Section 2.1.3. To group morphological properties of pellets, they are assigned
to the following classes: size, shape, compactness, morphology number, and spatial distri-
bution of hyphal material. In addition, micromorphological descriptors for dispersed hyphae
4
exist. However, due to a lack of suitable methods, these descriptors have not yet been studied
for pellets. Table 1 gives an overview of morphological properties and their corresponding
measurement techniques, which are elaborated in the following.
Size The size of individual pellets is specified by the projected area, various diameters,
perimeter, chord length, and signal length. Additionally, size distributions of pellet popula-
tions can be investigated.
The projected area is the two-dimensional area of the projection of a three-dimensional
object on a plane and often applied to pellets (Cairns et al., 2019a; Walisko et al., 2017;
Wucherpfennig et al., 2011). To determine the projected area, two-dimensional images are
acquired and analyzed. Based on the projected area, different diameters can be calculated.
The Feret diameter is defined as the distance between two parallel lines touching the edge of
the projected area (Cairns et al., 2019a; Walisko et al., 2017; Wucherpfennig et al., 2011). Ad-
ditionally, the surface equivalent spherical diameter of the projected area can be determined
(Schrinner et al., 2020). Some studies investigated the perimeter, defined as the total length of
the object boundary. A distinction is made between the perimeter of the convex area (Cairns
et al., 2019a; Wucherpfennig et al., 2011) and the perimeter of the projected area (Schrinner
et al., 2020; Walisko et al., 2017).
With focused beam reflectance measurement (FBRM) (Grimm et al., 2004; Kelly et al.,
2006; Lin et al., 2008; Pearson et al., 2003, 2004) and novel flow cytometry systems (Ehgart-
ner et al., 2017; Schrinner et al., 2020; Veiter and Herwig, 2019), the chord length and signal
length of pellets can be investigated, respectively. While the previously mentioned techniques
enable the investigation of numerous pellets individually, the application of laser diffraction
ultimately results in a size distribution of pellets (Lin et al., 2010; Petersen et al., 2008; Quin-
tanilla et al., 2018; Rønnest et al., 2012; Wucherpfennig et al., 2011).
Shape Based on image analyses of the projected area, the shape parameters circularity and
aspect ratio can be investigated. Schrinner et al. (2020) and Wucherpfennig et al. (2011)
applied the circularity, which is a function of the projected area and the perimeter:
Circularity = 4πPro jected area
Perimeter2 . (2.1)
While the projection of a perfect sphere would have a circularity of 1, values closer to 0
indicate elongated and/or non-convex objects. The aspect ratio is defined as the ratio between
the maximum and minimum Feret diameter (Cairns et al., 2019a; Wucherpfennig et al., 2011).
Symmetrical objects in all axis such as circles or squares would have an aspect ratio of 1,
whereas elongated objects result in higher aspect ratios. Thus, the aspect ratio is applied to
describe the elongation of pellets (Cairns et al., 2019a; Wucherpfennig et al., 2011).
Compactness The solidity is described as a surface property, measured on base of image
analysis, and defined as the ratio between the projected area and the convex area (Cairns
et al., 2019a; Wucherpfennig et al., 2011). A convex shape would result in a solidity of 1,
whereas non-convex shapes would result in lower values. Based on forward scatter signals
of flow cytometry, the properties relative annular diameter (RAD) and compactness can be
determined (Ehgartner et al., 2017). RAD describes the ratio of the loose pellet periphery to
the whole pellet whereas compactness represents the uniformity of the density of the pellet
core.
Combination of size, shape, and compactness Based on a combination of several
properties determined from image analysis, the dimensionless morphology number can be
investigated (Wucherpfennig et al., 2011):
morphology number =2√
A S√π D E
, (2.2)
where A is the projected area, S the solidity, D the maximum Feret diameter, and E the
elongation (aspect ratio). Perfectly round and convex pellets would result in a morphology
number of 1, whereas elongated and/or non-convex pellets would result in lower morphology
numbers.
Spatial distribution of hyphal material An established method to visualize the interior
of pellets is to acquire images of slices. In brief, pellets are frozen in embedding medium
and then cut into slices with a thickness of about 50 - 100 µm. Images of the slices are taken
either with light (Lin et al., 2010; Priegnitz et al., 2012) or confocal laser scanning (Hille
et al., 2005, 2009) microscopy. Lin et al. (2010) and Priegnitz et al. (2012) described the
spatial distribution of hyphal material only qualitatively by the appearance of dense and loose
regions in the pellet. Contrary, Hille et al. (2005, 2009) applied quantitative descriptors. They
investigated the hyphal fraction, i.e., the ratio of the volume of hyphae to the total volume, as
a function of the pellet radius.
Micromorphology of dispersed hyphae Because hyphae of more complex structures
superimpose, 2D image analysis can be only applied to investigate the micromorphology of
mycelia with a few branches. One important property is the total hyphal length, which can
be investigated manually with the help of measurement tools implemented in image analysis
softwares (Bocking et al., 1999; Kwon et al., 2013). Additionally, it can be determined auto-
matically by the investigation of the mycelial skeleton (Figure 2), which is the centerline of
hyphae with a thickness of one pixel (Barry et al., 2015; Cardini et al., 2020; Lecault et al.,
2007; Sachs et al., 2019; Schmideder et al., 2018; Vidal-Diez de Ulzurrun et al., 2019). Each
pixel of the skeleton can be assigned to be a tip (one neighbor), hyphae (two neighbors), or
branch (three neighbors). Four neighboring pixels would be assigned to the junction of two
overlapping hyphae. As shown in Figure 2, the number of tips and number of branches can
be investigated automatically based on analyses of the skeleton (Barry et al., 2015; Lecault
et al., 2007; Sachs et al., 2019; Vidal-Diez de Ulzurrun et al., 2019). However, the number of
tips and number of branches are often counted manually (Bocking et al., 1999; Kwon et al.,
2013). Dividing the total hyphal length by the number of tips results in the hyphal growth unit
7
(HGU) (Barry et al., 2015; Bocking et al., 1999; Choy et al., 2011; Colin et al., 2013; Kwon
et al., 2013; Sachs et al., 2019). Similar to the HGU, the internodal length (distance between
two branches) can be determined as a measure for the branching frequency (Du et al., 2016;
Lehmann et al., 2019; Sachs et al., 2019). Further, the hyphal diameter (Choy et al., 2011;
Colin et al., 2013; Lehmann et al., 2019) and the branch angle (Du et al., 2016; Lehmann
et al., 2019; Yang et al., 1992b) can be investigated.
Figure 2: Analysis of morphological development of Aspergillus niger mycelia: black marksthe centerlines of hypha, green the tips, and blue the branches. A1 - A4) Development ofmycelium A. B1 - B4) Development of mycelium B. In A5 and B5, the analyzed centerlinesof A4 and B4 are illustrated along with their respective microscopic image. Image analysisprocedure was adapted from Schmideder et al. (2018).
2.1.3 Applicability of morphological measurement techniques
Usually, microscopy is applied to determine the morphology of pellets. While, FBRM, laser
diffraction, and flow cytometry have become fast alternatives to microscopy, they often lead
to less detailed information. In the following, the advantages and limitations of the mentioned
measurement techniques are elaborated.
Light microscopy The state of the art to characterize the morphology is light microscopy
(Papagianni, 2014). Today, several open source image analysis tools simplify the offline-
analysis of pellets (Barry et al., 2015; Willemse et al., 2018) and dispersed hyphae (Barry
et al., 2015; Brunk et al., 2018; Cardini et al., 2020; Sachs et al., 2019; Vidal-Diez de Ulzurrun
et al., 2019). Further, the germination of spores (Brunk et al., 2018) and the tip extension and
branch formation of dispersed hyphae (Schmideder et al., 2018) can be tracked time-resolved
when the spores are fixed in a growth chamber (Figure 2). A combination of these tools,
the ubiquitous presence of light microscopes in laboratories, the opportunity to study diverse
morphological properties (compare Table 1), and the possibility to investigate hundreds of
pellets per time point (Schrinner et al., 2020) will further drive the use of light microscopy
to study filamentous microorganisms. However, the superimposition of hyphae hinders the
straightforward analysis of the micromorphology of more complex structures such as pellets.
to visualize the 3D morphology. However, CLSM requires a fluorescent signal and is limited
to about 50 - 150 µm in penetration depth. Thus, only the periphery can be visualized without
cutting the pellets (Hille et al., 2005, 2009; Villena et al., 2010). For unknown reasons, exist-
ing studies lack subsequent image analysis to investigate the micromorphology of the pellet
periphery. To determine the spatial distribution of the hyphal fraction inside pellets, Hille
et al. (2005, 2009) analyzed CLSM measurements of slices. However, the determined hyphal
fraction did not reflect the reality, since they used overlays of the acquired z-stacks. Thus, the
hyphal fraction was often 100 %, which is impossible even for densest packing. To the au-
thor’s knowledge there is no approach that analyzes the three-dimensional micromorphology
based on CLSM measurements.
Focused beam reflectance measurement FBRM enables the analysis of the chord
length and concentration of spores, spore agglomerates, and pellets offline (Pearson et al.,
2003, 2004; Whelan et al., 2012) as well as inline (Grimm et al., 2004; Kelly et al., 2006; Lin
et al., 2008). Advantages of FBRM are the inline applicability and the potential to measure
large quantities of objects. However, information about the shape and periphery of pellets are
missing and data interpretation remains challenging.
Laser diffraction To investigate the size distribution of spore agglomerates and pellets,
laser diffraction can be applied (Lin et al., 2010; Petersen et al., 2008; Quintanilla et al., 2018;
Rønnest et al., 2012; Wucherpfennig et al., 2011). Based on Fraunhofer diffraction theory,
the diffraction pattern of a laser beam can be analyzed to calculate the volumetric pellet size
distribution that best matches the measured pattern (Lin et al., 2010). Wucherpfennig et al.
(2011) applied laser diffraction in a bypass, whereas the other authors took samples before the
measurements. Laser diffraction has the advantage to be much faster than more commonly
used microscopy (Lin et al., 2010; Petersen et al., 2008; Quintanilla et al., 2018; Rønnest
et al., 2012). However, it fails to report properties that describe the shape and periphery of
pellets (Petersen et al., 2008; Wucherpfennig et al., 2011) as well as the pellet concentration
(Lin et al., 2010). According to Rønnest et al. (2012), the analysis is based on a number of
assumptions. For example, the shape of filamentous structures has to be assumed. However,
the authors suggested that validation studies could contribute to a reliable technique to analyze
size distributions of filamentous microorganisms.
Flow cytometry Ehgartner et al. (2017) described flow cytometry as a fast alternative to
microscopy. Further, they envision an online application through automated sampling sys-
tems. Similar to image analysis, flow cytometry allows the quantification of the pellet size
and the identification of the hairy region of the pellet periphery (Ehgartner et al., 2017; Schrin-
ner et al., 2020; Veiter and Herwig, 2019). In addition, flow cytometry reveals a compactness
parameter of pellets. However, information about the projected area is lacking and size ex-
clusion of samples occurs, which can result in an over-representation of small pellets (Veiter
and Herwig, 2019).
9
2.2 Interplay between morphology, transport of nutrients, and
metabolic activity of pellets
The interplay between morphology, transport of nutrients, and metabolic activity is a key
aspect for the productivity of filamentous pellets (Papagianni, 2004; Veiter et al., 2018; Zac-
chetti et al., 2018). In general, the optimal morphology varies with the desired product and
cannot be generalized (Gibbs et al., 2000; Krull et al., 2010). For example, the pellet form of
Aspergillus niger is used to produce citric acid, whereas its dispersed form serves for the pro-
duction of enzymes (Meyer et al., 2016). While pellets have some advantages over dispersed
hyphae (Section 1), limited nutrient availability (Cairns et al., 2019b; Krull et al., 2013) might
occur within pellets. For example, the concentration of oxygen is known to decrease towards
the center of pellets (Hille et al., 2005, 2009). This leads to a reduced growth and metabolic
activity in the core (Cairns et al., 2019b; Veiter et al., 2018) and can limit the production of
enzymes (Driouch et al., 2010; Veiter et al., 2018). However, reduced metabolic activity in
the center of pellets can also increase the production of secondary metabolites (Cairns et al.,
2019b; Veiter et al., 2018) such as Penicillin (Cronenberg et al., 1994). This demonstrates
that a detailed understanding of the profile of nutrients and the resulting metabolic activity in
pellets is of crucial importance.
2.2.1 Spatial distribution and mass transport of nutrients inside pellets
The spatial distribution of nutrients inside pellets is determined by metabolic rates and trans-
port processes through their dense structure (Celler et al., 2012; Cui et al., 1998). In the
following section, the focus is on the investigation of nutrient profiles in pellets and mass
transport properties of hyphal networks.
Profile of nutrients In general, studies measuring the profile of nutrients are rare. A rea-
son might be the need of complicated experimental setups, where pellets are fixed in defined
chambers before the concentration profile of oxygen or glucose can be measured with micro-
electrodes (Cronenberg et al., 1994; Hille et al., 2005, 2009; Wittier et al., 1986). To guarantee
a realistic nutrient transport and keep the morphology intact, the handling of the pellets and
microelectrodes as well as the design and control of the chambers are challenging. For exam-
ple, Hille et al. (2005, 2009) applied a thin microelectrode with a tip diameter similar to the
diameter of hyphae.
Experiments unveiled oxygen as prime limiting nutrient in fungal pellets (Cronenberg et al.,
1994; Veiter et al., 2018). Measurements of the oxygen profile in pellets greater than 1 mm
revealed a restricted availability of oxygen with only the outer 100 - 300 µm being supplied
(Cronenberg et al., 1994; Hille et al., 2005, 2009; Wittier et al., 1986). In addition to oxygen,
Cronenberg et al. (1994) determined the concentration profile of glucose. Although glucose
penetrated Penicillium chrysogenum pellets at early cultivation stages almost at bulk level, no
glucose consumption was determined in their core. Since fungal metabolism requires oxygen,
this was explained by the measured absence of oxygen in the core. At late cultivation stages,
10
the pellets were prone to fragmentation and autolysis and showed a decreased metabolic ac-
tivity in the core. While pellets were completely penetrated by oxygen and glucose, glucose
consumption still only occurred in the periphery. The authors explained the lost metabolic ac-
tivity in the core by the irreversible inhibition in early cultivation stages caused by the absence
of oxygen.
Especially Cronenberg et al. (1994) demonstrated the interplay between the profile of nu-
trients and the metabolic activity in pellets. Further, all mentioned studies revealed that con-
centration profiles in pellets are highly affected by their structure. In addition to the size of
pellets, the density plays a decisive role. For example, Hille et al. (2005) observed a much
steeper decrease of the oxygen concentration in dense A. niger pellets. Moreover, measuring
the development of nutrient profiles in inactivated pellets is the only method to investigate
their diffusivity, which is be elaborated in the following.
Mass transport of nutrients Many authors suggest diffusion as the only mass transport
phenomenon of oxygen and nutrients inside pellets (Cui et al., 1998; King, 1998; Silva et al.,
2001). In addition to diffusive transport, convective transport is proposed to contribute to the
nutrient-supply of pellets with a loose periphery, especially if they are prone to turbulent flow
regimes (Cronenberg et al., 1994; Hille et al., 2009). Because diffusive mass transport was
shown to be the most important transport phenomenon inside pellets, its theoretical back-
ground is elaborated in the following. The diffusivity of component i is described with the
effective diffusion coefficient Di,eff (Becker et al., 2011):
Di,eff = Di,bulk · keff , (2.3)
where Di,bulk is the diffusion coefficient in the pure bulk medium and keff determines the
reduction to the effective diffusion coefficient. While the effective diffusion factor keff is
dependent on the geometry of the pores, it is independent of the diffusing substance.
Besides measurement of nutrient profiles, Cronenberg et al. (1994) and Hille et al. (2009)
investigated the effective diffusivity of P. chrysogenum and A. niger pellets, respectively. For
this purpose, Hille et al. (2009) placed one microelectrode each at the periphery and in a de-
fined depth of inactivated pellets. The inactivation of the pellets prevented the consumption
of oxygen. After saturating the medium with nitrogen, the measurement chamber was aerated
with pure oxygen and the change of the concentration was recorded by both microelectrodes.
Based on the development of the concentrations and Fick’s second law, they were able to
fit the effective diffusion coefficient. Similar to Hille et al. (2009), Cronenberg et al. (1994)
determined the development of the oxygen or glucose concentration by stimulus response ex-
periments inside inactivated pellets to further estimate the effective diffusivity. As expected,
dense pellet peripheries resulted in low effective diffusivities. The minimum effective diffu-
sion factor ke f f was about 0.4 and 0.6 in the studies of Hille et al. (2009) and Cronenberg
et al. (1994), respectively. As Cronenberg et al. (1994) described the morphology of pellet
slices only qualitatively, they were not able to correlate the morphology with the effective
diffusivity. Contrary, Hille et al. (2009) quantified the morphology of pellet slices with the
11
radial profile of the hyphal fraction. However, as mentioned in Section 2.1.3, they missed the
three-dimensional information of the hyphal network. Thus, hyphae superimposed in two-
dimensional projections of slices and the hyphal fraction was often 100 %. That is impossible
even for densest packing. Nevertheless, they correlated the effective diffusivity with the local
derivation of the hyphal fraction.
To the author’s knowledge, Hille et al. (2009) was the only experimental approach to
correlate the morphology and diffusivity of hyphal networks. Although studies about fi-
brous materials revealed that porosity and tortuosity are important drivers of ke f f (Tomadakis
and Robertson, 2005; Vignoles et al., 2007), modeling approaches of filamentous pellets
(Buschulte, 1992; Cui et al., 1998; Lejeune and Baron, 1997; Meyerhoff et al., 1995) ei-
ther neglect the tortuosity of the hyphal network or assume a constant value to model the
diffusivity of nutrients. A detailed list of applied and determined correlations between the
morphology and effective diffusion factors of filamentous microorganisms and fibrous mate-
rials can be found in Paper II, Table 1.
2.2.2 Metabolic activity
The limited availability of nutrients in central parts of pellets can lead to a physiological
heterogeneity inside pellets (Zacchetti et al., 2018) and to a reduced growth and growth-
related metabolism (Hille et al., 2009; Zhang and Zhang, 2016). Metabolic activity can be
identified by staining and analyzing active as well as inactive regions. In this way, many
studies determined metabolic inactive cores and active shell layers of pellets. Staining as
indicator of metabolic activity is conducted either with chemicals (Bizukojc and Ledakowicz,
2010; Nieminen et al., 2013; Schrinner et al., 2020; Veiter and Herwig, 2019) or by using
fluorescence proteins expressed by the host organism (Driouch et al., 2012, 2010; Tegelaar
et al., 2020).
The author of this dissertation contributed to a cooperative study about the metabolic ac-
tivity in filamentous bacterial pellets (Schrinner et al., 2020), which is not embedded in the
results section of this thesis. In this study, active and inactive regions of Lentzea aerocoloni-
genes pellets were stained with SYTO9 (green fluorescence through intercalation with DNA
of predominantly intact and active cells) and propidium iodide (red fluorescence as result from
DNA intercalation in inactive cells with compromised membranes), respectively. As illus-
trated in Figure 3, metabolically different regions of pellet slices were distinguished through
the analysis of CLSM images. Because pellet slicing followed by image analysis is of high
manual effort, flow cytometry was applied to detect active and inactive fractions of hundreds
of stained pellets. While flow cytometry enabled the analysis of a statistically relevant number
of pellets, image analysis of pellet slices provided shape information.
Fluorescence staining of active and inactive pellet regions can also be applied to filamen-
tous fungi. Similar to the mentioned study about filamentous bacteria (Schrinner et al., 2020),
Veiter and Herwig (2019) analyzed P. chrysogenum pellets based on CLSM and flow cytom-
etry. Bizukojc and Ledakowicz (2010) stained growing regions of Aspergillus terreus with
lactophenol methyl blue. This staining procedure allows to analyze growing regions through
12
Figure 3: Determination of living (green), dead (red) and a combination of living and dead(yellow) pellet areas. Image analysis was performed based on CLSM measurements of stainedLentzea aerocolonigenes pellet slices. (a) Original image with green and red fluorescent areas(black circles are air bubbles that are occasionally enclosed in the sectioning medium). (b)Separation in green fluorescence and red fluorescence. (c) Processed images. (d) Imagewith living, dead, and a combination of living and dead pellet areas. Figure is from jointpublication (Schrinner et al., 2020).
light microscopy of pellet slices. Further, the authors correlated the fraction of active growing
regions with the formation of the desired product lovastatin. Instead of activity staining with
chemicals, Driouch et al. (2012, 2010) engineered genetically modified A. niger strains that
co-express green fluorescent protein (GFP) together with the desired product glucoamylase
or fructofuranosidase, respectively. Thus, they were able to quantify metabolic active regions
based on CLSM images of pellet slices. In Driouch et al. (2012), large pellets were only
active in a 200 µm surface layer. The authors concluded that the inner region of large pellets
does not contribute to the production of the desired enzymes, which is probably caused by
diffusion limitation of oxygen or other nutrients.
In several studies, the volume fraction of the active shell layer and the formation of desired
products increased with decreasing pellet diameter (Bizukojc and Ledakowicz, 2010; Dri-
ouch et al., 2012, 2010; Tegelaar et al., 2020). However, a correlation between the metabolic
activity and the detailed morphology within pellets, e.g., the spatial distribution of hyphal
material and the number of (active) tips, is lacking. Additionally, there is no study to date that
determines both the nutrient profile and metabolic heterogeneity within pellets.
13
2.3 Morphological engineering
There is no doubt that the morphology of filamentous microorganisms strongly affects their
productivity in submerged cultures. Thus, the precise control of the morphology, i.e., morpho-
logical engineering, is of crucial importance (Krull et al., 2013). In this Section, experimental
and modeling approaches are described to engineer the morphology of filamentous pellets.
2.3.1 Experimental approaches
In this thesis, it is distinguished between genetic and process engineering approaches to alter
the morphology.
Genetics Two recent reviews highlight genetic aspects that impact the morphogenesis of
filamentous fungi (Cairns et al., 2019b; Zhang and Zhang, 2016). In the following, a few
exemplary filamentous fungal characteristics are described that can be specifically modified
to allow targeted strain development.
As stated earlier, spore agglomeration is one of the driving factors for the formation of pel-
lets. One known factor for spore agglomeration is their hydrophobicity (Zhang and Zhang,
2016). The absence of spore cell wall associated hydrophobins DewA and RodA in As-
pergillus nidulans knockout mutants reduced the ratio of biomass present as pellets. Further,
the average size of pellets decreased (Dynesen and Nielsen, 2003). Fontaine et al. (2010)
showed that cell wall 1-3glucans become exposed at the cell surface during spore swelling and
induce the agglomeration of germinating Aspergillus fumigatus conidia. Additionally, spores
of species with 1-3glucan synthase gene (ags) in their genome (A. fumigatus and P. chryso-
genum) agglomerate, whereas no spore agglomeration can be observed for organisms without
ags (R. oryzae and Trichoderma reesii). Further, A. niger alb1 knockout mutants are deficient
in melanin biosynthesis and showed altered surface structures and charge of spores (Priegnitz
et al., 2012). The authors concluded that spore agglomeration differs between the wild type
and the mutant in pH-dependent manner due to the changed surface charge.
Besides spore agglomeration, the branching frequency has a strong influence on the mor-
phological development (Cairns et al., 2019b). As shown by several authors, the branching
frequency can be genetically modified (Biesebeke et al., 2005; Fiedler et al., 2018; He et al.,
2016; Kwon et al., 2013). Fiedler et al. (2018) reported more compact macromorphologies
of the hyperbranching phenotype. In addition to the change in morphology, productivity can
also be altered by the branching frequency. Especially for the production of enzymes, a high
tip to biomass ratio seems to be beneficial (Biesebeke et al., 2005; He et al., 2016). However,
in some cases an elevated tip to biomass ratio cannot be correlated with an increased protein
titer (Cairns et al., 2019b).
The described genetic modifications prove that the morphology can be influenced specifi-
cally by molecular biological approaches. To investigate the impact of genetic approaches on
the formation of pellets and the pellet size, morphological measurement techniques shown in
Table 1 can be applied. However, the influence on the micromorphology of pellets can of-
14
ten not be determined. For example, hyperbranching can be quantified for dispersed hyphae
based on image analysis, while such methods do not exist for pellets so far.
Process engineering Several review articles describe process engineering approaches for
altering the morphology of filamentous fungal pellets (Böl et al., 2021; Krull et al., 2010,
2013; Veiter et al., 2018). Due to space limitation, only a small overview is presented about
approaches concerning inoculation, medium composition, and fluid dynamics.
Inoculation strategies are known for their high impact on the morphological development
(Prosser and Tough, 1991; Veiter et al., 2018). Papagianni and Mattey (2006), e.g., inocu-
lated bioreactors with A. niger spores in concentrations ranging from 104 to 109 spores per
milliliter and observed a clear transition from pelleted to dispersed macromorphologies. Con-
trary to A. niger (coagulative pellet formation type), non-coagulative pellet formation type
R. stolonifer (synonym R. nigricans) develops larger pellets at lower spore concentrations
(Žnidaršic et al., 2000). Besides inoculation with spores, inoculation with pellets can be a
promising strategy for A. niger cultivations (Wang et al., 2017).
The composition of the medium offers numerous opportunities to influence the morphol-
ogy. Besides traditional approaches such as the variation of pH (Priegnitz et al., 2012) and
nutrient-sources (Papagianni et al., 1999), two approaches are currently coming into focus:
the adjustment of the osmolality (Böl et al., 2021; Wucherpfennig et al., 2011) and the addi-
tion of particles (Böl et al., 2021; Karahalil et al., 2019). Wucherpfennig et al. (2011) showed
that A. niger pellets decrease in size with increasing osmolality. The addition of micropar-
ticles (diameter mostly < 50 µm (Böl et al., 2021)) can also reduce the size of filamentous
pellets and increase enzyme production (Antecka et al., 2016; Karahalil et al., 2019).
According to Krull et al. (2013), the reactor geometry, stirrer shape and size, and dissipated
energy are among the fluid dynamic-related criteria affecting the morphology and productiv-
ity in stirred tank reactors (STR). They concluded that low power input leads to inadequate
nutrient distribution and gas dispersion, whereas high power input can result in cell wall dam-
age. Consequently, high mechanical forces can result in chipping off hyphae from pellets (Cui
et al., 1997), which might reseed the fermentation broth. The following studies show that a
more profound understanding is required to correlate the power input with the morphology.
On the one hand, A. niger can grow to large pellets for low stirrer speeds, whereas high stir-
rer speeds can result in dispersed hyphae (El-Enshasy et al., 2006). On the other hand, an
increased mechanical power input can increase the density of the pellet periphery of A. niger
(Lin et al., 2010). The authors concluded that the increased density of the pellet periphery
might limit the mass transport of nutrients. Compared to STR, other reactor types such as air
lift column reactors (Gibbs et al., 2000; Xin et al., 2012) and rocking motion reactors (RMR)
(Kurt et al., 2018) can reduce shear stress. According to Kurt et al. (2018), RMR can result
in higher growth rates and more pelletized A. niger structures than STR.
The mentioned process engineering approaches offer numerous opportunities for targeted
morphological engineering. In particular, they enable the cultivation of predominantly pel-
letized or dispersed morphologies and affect the size of pellets. Some studies even showed the
15
possibility to alter the density of the pellet periphery, which could contribute as a diffusion
barrier. However, the impact of process engineering approaches on the micromorphology
within pellets has not yet been examined, which is caused by a lack of appropriate methods.
2.3.2 Modeling approaches
Several authors emphasize the demand for mechanistic modeling in order to predict the mor-
phological development for altered cultivation conditions, to reduce elaborate empirical tests,
and to improve the fundamental understanding between morphology and productivity (Celler
et al., 2012; Grimm et al., 2005; King, 1998; Posch et al., 2013; Veiter et al., 2018). In this
thesis, morphological modeling of filamentous pellets is categorized in 1) microscopic ap-
proaches, 2) continuum approaches, and 3) population balance modeling (PBM) approaches.
Microscopic approaches describe the growth of pellets, considering all positions of hyphae,
branches, and tips. Contrary, continuum approaches are less detailed and outline the ra-
dial distributions of morphological properties in pellets. While microscopic and continuum
approaches predict the development of individual pellets, PBM considers the pellet hetero-
geneity within a cultivation. As this thesis is based on the morphological characterization of
individual pellets, PBMs are not elaborated here. However, Section 6.3 includes a discussion
on how such modeling approaches could benefit from the methods and findings of this thesis.
Microscopic approaches As shown in Figure 4, microscopic models describe the devel-
opment of hyphal networks including the location of hyphae, tips, and branches. While two-
dimensional approaches consider the growth of filamentous microorganisms on a plane, three-
dimensional ones enable the simulation of hyphal networks in submerged cultures. Since this
thesis is about filamentous pellets, only three-dimensional models that incorporate the simu-
lation of whole pellets are considered here.
Based on stochastic rules for tip extension and branching, pellets can be simulated starting
from a single spore (Celler et al., 2012; Lejeune and Baron, 1997; Meyerhoff et al., 1995;
Yang et al., 1992a). For this, microscopic approaches considered following intracellular phe-
nomena: septation of hyphae (Celler et al., 2012; Yang et al., 1992a), distinction into apical,
subapical, and hyphal compartments (Celler et al., 2012), and consumption and flow of a com-
ponent in hyphae that influences the tip extension rate (Yang et al., 1992a). Recent advances in
measuring and modeling the intracellular transport of secretory vesicles, which transport cell
wall components to the hyphal tips (King, 2015; Kunz et al., 2020), have not been included
in three-dimensional microscopic pellet models yet. Contrary, several approaches considered
the diffusive mass transport of nutrients from the cultivation broth through the hyphal network
of pellets (Celler et al., 2012; Lejeune and Baron, 1997; Meyerhoff et al., 1995). For this, the
authors assumed spherical symmetry and computed the transport and consumption of nutri-
ents based on partial differential equations. To prevent hyphae from growing into each other,
collision detection can be included (Celler et al., 2012). Another phenomenon that can be
considered in microscopic models is the abrasion of pellets due to shear forces (Celler et al.,
2012; Meyerhoff et al., 1995). However, the development of coagulative pellets has not been
16
addressed sufficiently yet. While only Lejeune and Baron (1997) considered the development
of pellets from spore agglomerates, these simulations were limited to 100 randomly placed
spores. Hence, this was not realistic for coagulative pellet formation.
Most breakthroughs of microscopic approaches were achieved in the nineties including
three-dimensional representation, intracellular processes, diffusive mass transport of nutri-
ents into pellets, and abrasion of pellets. However, stagnation in model development can be
observed. That might be caused by the absence of experimental data for model validation.
Especially the hyphal network of the pellet interior cannot be quantified sufficiently to com-
pare simulated and cultivated pellets (Sections 2.1.2 and 2.1.3). Additionally, the models lack
well-founded correlations between the structure and the diffusivity for the computation of the
nutrient supply (Section 2.2.1).
Figure 4: Morphological modeling of pellets. Microscopic approach: own work based onCeller et al. (2012); Continuum approach: own work based on Buschulte (1992).
Continuum approaches First continuum approaches to model pellet growth were based
on the cube root relation for the development of the fungal biomass concentration in culti-
vations (Emerson, 1950; Pirt, 1966). The cube root relation is based on the hypothesis that
pellet growth occurs only in the periphery. Contrary, the pellet center is assumed to consist
of non-growing mycelium, into which oxygen does not diffuse. In later studies, the oxygen-
limitation inside pellets was confirmed experimentally (Cronenberg et al., 1994; Hille et al.,
2005; Wittier et al., 1986). Additionally, image analysis of pellet slices revealed the spatial
heterogeneity of the hyphal material inside pellets qualitatively (Hille et al., 2005; Lin et al.,
17
2010; Priegnitz et al., 2012). However, the cube root relation does not consider spatial hetero-
geneities of the hyphal fraction inside pellets. Buschulte (1992) and Meyerhoff et al. (1995)
overcame this limitation. Their approaches assumed spherical symmetry of pellets concern-
ing the spatial distribution of hyphal material and nutrients (Figure 4). Similar to microscopic
approaches (Celler et al., 2012; Lejeune and Baron, 1997), transport and consumption of
oxygen inside pellets was computed based on partial differential equations. In this way, the
inactivation of mycelial growth in the pellet center can be addressed based on the nutrient
profile. According to King (1998), a description with continuous variables for the biomass,
such as concentrations of hyphae and tips seems meaningful. The approach introduced by
Buschulte (1992) considers the radial concentrations of hyphae, tips, oxygen, and other sub-
strates inside pellets during their growth and is based on coupled partial differential equations
for these concentrations. With a so called layer model, Meyerhoff et al. (1995) described a
simplification of the approach of Buschulte (1992). This approach divides the pellet in a few
spherical shells and assumes constant values for hyphal fraction, tips, and nutrients inside
these shells. To reduce computational time, this model intentionally omits partial differential
equations except for the transport of nutrients into the pellet. Compared to a microscopic
model similar to Yang et al. (1992a), the continuum approach resulted in an about 60 - 100
fold reduction of the demands for computing capacity (Meyerhoff et al., 1995).
While the computational time of continuum approaches is lower compared to microscopic
ones, the level of detail remains high. Especially the morphological development of spherical
symmetric pellets can be well described. Although continuum approaches seem promising
to optimize filamentous fungal bioprocesses, recent advances are missing. To the author’s
knowledge, the model of Buschulte (1992) is still the most detailed approach. Similar to
microscopic approaches, continuum ones lack sufficient validation procedures, which might
be the reason for the stagnation in model development. Validation procedures would require
radial profiles of morphological properties and substrates and a well-founded law for the
diffusive mass transport through filamentous fungal networks. However, these properties are
difficult or even impossible to measure with current techniques (Sections 2.1.2, 2.1.3, and
2.2.1).
18
3 Problem definition
Although pellets are often exploited in industrial processes using filamentous fungi, their
inner structure and the resulting diffusion barrier for nutrients, oxygen, and secreted metabo-
lites remained largely unexplored. The following knowledge gaps regarding the micromor-
phology and diffusivity of filamentous fungal pellets were identified based on the theoretical
background.
Method to analyze the micromorphology of whole intact pellets does not exist.
The micromorphology within pellets is known to have a high impact on the productivity
of bioprocesses using filamentous fungi. Although many methods have been developed to in-
vestigate the macromorphology and periphery of pellets, there is no non-destructive approach
to visualize the three-dimensional micromorphology of whole pellets. Further, no method ex-
ists for the analysis of 3D images of filamentous fungal networks. Methods to visualize and
analyze the micromorphology of whole intact pellets would open new paths towards morpho-
logical engineering. For example, the impact of genetic and process engineering approaches
on the morphology of pellets could be investigated in utmost detail. Additionally, existing
morphological modeling approaches without suitable micromorphological input from exper-
iments could be validated and improved.
Correlation between the three-dimensional structure and the diffusivity of pellets is
not described. The supply with nutrients and oxygen is crucial for the metabolic activity
and product formation in pellets. Studies revealed that the transport of nutrients and oxygen
through dense hyphal networks is mainly driven by diffusion. However, a well-founded cor-
relation between the micromorphology and the diffusivity through fungal pellets is lacking in
literature. Such a correlation would enable the prediction of the diffusive transport of nutri-
ents, oxygen, and secreted metabolites in filamentous fungal pellets and thus, contribute to
the targeted design of pellet morphologies.
19
4 Methods for problem solving
The problem definition illustrates that new methods are required to reveal the micromor-
phology and diffusivity of filamentous pellets. Therefore, the following methods have been
developed. A detailed description of all methods can be found in the embedded papers.
Method to visualize whole intact pellets three-dimensionally. Applying X-ray mi-
crocomputed tomography (µCT) measurements, it became feasible to visualize the three-
dimensional (3D) network of hyphae forming filamentous fungal pellets. While this technique
enabled the non-destructive visualization of whole pellets with several hundred micrometers
in diameter, the hyphae with diameters down to 3 µm were resolved.
Method to analyze three-dimensional images of pellets. To quantify the micromor-
phology of whole pellets, automated image analysis was conducted on the acquired µCT
images. After binarization of the hyphal network, the centerlines of hyphae were determined
and used to locate tips and branches. This procedure enabled the investigation of morpholog-
ical properties such as the hyphal length, number of tips, number of branches, hyphal growth
unit (HGU), and porosity of whole pellets. Additionally, the spatial distributions of the hyphal
fraction, tip density, and branch density were determined.
Method to correlate structure and diffusivity of pellets. Diffusion computations were
conducted through numerous representative sub-volumes per µCT measured pellet. The com-
putations resulted in a diffusivity for each sub-volume, which is a measure for the geometrical
diffusion hindrance and independent of the diffusing substance. Correlation analysis between
the diffusivities and the structures of several hundred Aspergillus niger sub-volumes unveiled
a diffusion-law with respect to the solid hyphal fraction.
Method to determine a universal law for the diffusion through mycelial networks.
To consider the broad morphological range of filamentous fungi, the structures and diffu-
sivities of both µCT measured and simulated pellets were correlated based on the method
described above. While µCT measured pellets from four fungal species already showed
strongly differing morphologies, Monte Carlo simulated pellets covered the broad morpho-
logical range of filamentous microorganisms. To obtain the required 3D structures of the
simulated pellets, an existing microscopic model was extended and implemented. Analysis
of all measured and simulated pellets unveiled a universal law for the diffusion of nutrients,
oxygen, and secreted metabolites with the solid hyphal fraction as the only independent vari-
able.
20
5 Results
5.1 Paper I: An X-ray microtomography-based method for
detailed analysis of the three-dimensional morphology of
fungal pellets (Schmideder et al., 2019a)
Summary
Although the micromorphology of filamentous fungal pellets is strongly linked to the produc-
tivity in bioprocesses, there is no method to visualize the detailed three-dimensional morphol-
ogy of whole intact pellets. Further, no method for the analysis of three-dimensional images
of filamentous fungal networks exists to date. To enable the non-destructive visualization of
whole pellets, we developed a protocol based on X-ray microcomputed tomography (µCT).
Exemplarily, we investigated pellets of Aspergillus niger and Penicillium chrysogenum. The
binarization of the images enabled the determination of the hyphal network within whole pel-
lets. Based on the binarized hyphal network, skeletonization was conducted to investigate the
location of tips and branches as well as the total hyphal length. Thus, the cumulative mor-
phological properties total hyphal length, total tip number, total branch number, porosity, and
hyphal growth unit of whole pellets can be determined. Additionally, multiple hypotheses
about the morphological development of fungal pellets can be drawn from the spatial distri-
butions of the hyphal fraction, tip density, and branch density. Based on µCT measurements
and image analysis, the outcome of experimental morphological engineering approaches on
pellet structures can be investigated in unprecedented detail. Further, the determined three-
dimensional morphology will serve as valuable input to validate and improve existing mor-
phological modeling approaches. As shown in Paper II and III, the structure of analyzed
pellets can also be used to compute the resulting diffusion barrier for nutrients, oxygen, and
secreted metabolites.
Author contributions
Stefan Schmideder did the conception and design of the study and wrote the manuscript,
which was edited and approved by all authors. Heiko Briesen and Vera Meyer supervised the
study. Lars Barthel and Ludwig Niessen cultivated and freeze-dried fungal pellets. Stefan
Schmideder and Michaela Thalhammer developed a protocol for µCT measurements. Stefan
Schmideder, Tiaan Friedrich, and Tijana Kovacevic developed 3D image analysis of µCT
measurements. Stefan Schmideder analyzed the results. Stefan Schmideder, Lars Barthel,
Heiko Briesen, and Vera Meyer interpreted the results.
21
Received: 14 August 2018 | Revised: 21 December 2018 | Accepted: 9 January 2019
DOI: 10.1002/bit.26956
AR T I C L E
An X‐ray microtomography‐based method for detailedanalysis of the three‐dimensional morphologyof fungal pellets
Stefan Schmideder1 | Lars Barthel2 | Tiaan Friedrich1 | Michaela Thalhammer1 |Tijana Kovačević1 | Ludwig Niessen3 | Vera Meyer2 | Heiko Briesen1
Award Numbers: BR 2035/11‐1, ME 2041/5‐1, DFG INST 95/1111‐1, BR 2035/11‐1 and
ME 2041/5‐1
Abstract
Filamentous fungi are widely used in the production of biotechnological compounds.
Since their morphology is strongly linked to productivity, it is a key parameter in
industrial biotechnology. However, identifying the morphological properties of
filamentous fungi is challenging. Owing to a lack of appropriate methods, the
detailed three‐dimensional morphology of filamentous pellets remains unexplored. In
the present study, we used state‐of‐the‐art X‐ray microtomography (µCT) to develop
a new method for detailed characterization of fungal pellets. µCT measurements were
performed using freeze‐dried pellets obtained from submerged cultivations. Three‐dimensional images were generated and analyzed to locate and quantify hyphal
material, tips, and branches. As a result, morphological properties including hyphal
length, tip number, branch number, hyphal growth unit, porosity, and hyphal average
diameter were ascertained. To validate the potential of the new method, two fungal
pellets were studied—one from Aspergillus niger and the other from Penicillium
chrysogenum. We show here that µCT analysis is a promising tool to study the three‐dimensional structure of pellet‐forming filamentous microorganisms in utmost detail.
The knowledge gained can be used to understand and thus optimize pellet structures
by means of appropriate process or genetic control in biotechnological applications.
Others assume spherical symmetry of pellets and consider morpho-
logical properties, such as the hyphal length density and tip density
dependent on the radius (Buschulte, 1992; Meyerhoff, Tiller,
& Bellgardt, 1995). However, these modeling approaches have not
been fully validated yet owing to the significant lack of knowledge of
three‐dimensional pellet morphologies.
This study thus introduces a new method for a detailed
investigation of the morphological properties of filamentous pellets
and is structured as follows: In Section 2, the preparation of fungal
pellets and the microtomography (µCT) measurements are described.
The applied µCT measurements offer a nondestructive way to
visualize complex morphologies and are applied intensively in
medical and biological research as well as material science (Salvo
et al., 2003; Stock, 2008). Since the subsequent data processing of
fungal pellets cannot be clearly separated into the methodological
development of the image processing and the results of this
processing, Section 3 is introduced. By applying three‐dimensional
image analysis, the locations of hyphae, tips, and branches were
determined. To demonstrate the potential of the described method,
two freeze‐dried fungal pellets from submerged cultivations from A.
niger and P. chrysogenum, respectively, were investigated in detail in
the Section 4. Based on the processed data of the fungal pellets,
morphological properties like the hyphal length, number of tips,
number of branches, hyphal growth unit (HGU), porosity, and
average diameter of hyphae were investigated.
2 | MATERIALS AND METHODS
2.1 | Pellet preparation
Conidiospores of P. chrysogenum strain MUM17.85 (Micoteca da
Universidade do Minho, Braga, Portugal) and A. niger strain MF22.4
(Fiedler, Barthel, Kubisch, Nai, & Meyer, 2018) were obtained from
agar plate cultures by using standard procedures for filamentous
fungi (Bennett & Lasure, 1991). The spores were grown in liquid
cultivation media for P. chrysogenum (yeast carbon base; Difco,
Franklin Lakes, NJ) and A. niger (complete medium; Meyer, Ram, &
Punt, 2010) for 24–48 hr until pelleted structures became visible.
Single pellets were then carefully removed by pipetting and were
washed three times with water. Samples were frozen in liquid
nitrogen while pellets were floating in water to preserve their
structure and were subsequently freeze‐dried. A FreeZone device
(Labcono, Kansas City, MO) was used to freeze dry the A. niger pellet
at 0.014 mbar and −55°C for 24 hr. The freeze‐drying of the P.
chrysogenum pellet was performed using a FreeZone 2,5 PLUS device
(Labconco) at 0.002 mbar and ambient temperature for 24 hr. To
prevent the freeze‐dried pellets from absorbing water, the samples
were stored in sealed Eppendorf tubes for further use. To investigate
the influence of the applied freeze‐drying process on the morphology
of filamentous pellets, we compared the same pellets in two states:
wet (immediately after fermentation) and freeze‐dried. Light micro-
scope images of wet and freeze‐dried A. niger and P. chrysogenum
pellets are shown in Figures S1–S5. After 48 hr fermentation of A.
niger and P. chrysogenum, the diameter of freeze‐dried pellets
decreased by 9% and 10% on average, respectively. The pellet
diameter has been calculated on base of the outermost hyphae of the
2 | SCHMIDEDER ET AL.
23
pellets. The fermentation of P. chrysogenum for 24 hr resulted in
looser pellets (Figure S3) and a decrease in the dried pellet diameter
of 13%. In addition, we measured the diameter of wet and freeze‐dried hyphae at the pellet periphery of P. chrysogenum pellets
manually with FIJI. The average diameter of wet hyphae was 3.6 µm,
whereas freeze‐dried hyphae had an average diameter of 3.5 µm. In
addition, the freeze‐dried samples do still exhibit hairy regions in the
outer part of the pellets (Figures S4 and S5).
2.2 | X‐ray microtomography
Three‐dimensional images of the fungal pellets were acquired using a
custom‐built X‐ray microtomography system (XCT‐1600HR; Matrix
Technologies, Feldkirchen, Germany). Between the open tube and
the detector, the fungal pellets were fixed on a sample holder, which
rotated during the measurement. The tube of the µCT system
generates a cone beam. Thus, two‐dimensional projections were
obtained from various angles. The projections were reconstructed
using a custom‐designed software (Matrix Technologies) that uses
CERA (Siemens, Munich, Germany) to receive three‐dimensional
images. Because freeze‐dried filamentous pellets have a low density,
low energy (60 kV, 25 µA) was used to generate the cone beam. The
three‐dimensional images had a resolution of 1 µm (i.e., the edge
length of the voxels was 1 µm). To fix the A. niger pellet on top of the
sample holder, an instant adhesive (UHU, Bühl, Germany) was used.
Partial embedding of the A. niger pellet in the instant adhesive
guaranteed structural stability. For the P. chrysogenum pellet, a
double‐sided tape (Tesa, Norderstedt, Germany) was sufficient to
guarantee structural stability.
3 | DATA PROCESSING
This section describes the detection of tips, branches, and hyphal
material in pellets by using image analysis. The analysis was
conducted with the µCT images of the A. niger (Figure 1) and the P.
chrysogenum pellet (Figure 2) acquired as described in Section 2.
3.1 | Preprocessing
Preprocessing aimed to generate binarized three‐dimensional images
with the hyphal material as foreground. Thereby, voxels originating
from noisy data and sample fixation materials must be separated
from the pellet.
Both fixation materials, the instant adhesive for A. niger and the
double‐sided adhesive tape for P. chrysogenum, showed gray values
similar to those of the pellets. Because of this, the automated
segmentation of the pellet turned out to be difficult to implement. To
achieve successful segmentation, the images were cropped. Part of
the A. niger pellet enclosed in the instant adhesive was deformed. For
further investigation, a conically shaped section was cropped from
the raw data using MATLAB (version R2016a; MathWorks, Natick,
MA), resulting in a structure not affected by the fixation material
(Figure 1a). This cone was used to represent the whole pellet.
According to visual observations, the tip of the cone was assumed to
be the center of the pellet. The opening angle of the cone was chosen
to be 65°. The P. chrysogenum pellet was cropped by cutting off a
small XY‐orientated slice at the bottom of the three‐dimensional
image with the software MAVI (version 1.4.1; Fraunhofer ITWM,
Kaiserslautern, Germany). After cropping, the 16‐bit grayscale
F IGURE 1 µCT image of A. niger pellet;two‐dimensional projections of a
three‐dimensional image. (a) Whole pelletwith XZ‐orientation; red object representsthe analyzed cone of the pellet;transparent white object represents the
remaining pellet and fixation material.(b–d) Slices (100 µm) of differentorientations; center of the slices is the
pellet center; image processed with basicrendering only; (b) XY‐orientation;(c) YZ‐orientation; (d) XZ‐orientation[Color figure can be viewed atwileyonlinelibrary.com]
SCHMIDEDER ET AL. | 3
24
images (i.e., voxels of µCT images exhibiting gray values between 0
and 65,535) were binarized to differentiate between the hyphae and
background (Figure 3a,b). The binarization was performed using
MAVI by setting a gray value threshold that was calculated using
Otsu’s method (Otsu, 1979). The voxels with gray values higher than
the gray value threshold were designated as hyphal material. Small
interconnected objects, which were not part of the pellet, were
eliminated. This was achieved by labeling neighboring voxels with
MAVI and deleting all but the largest object. It has to be mentioned
that MATLAB or FIJI could also be applied to the preprocessing steps
conducted in MAVI. However, the three‐dimensional rendering
performance of MAVI was highly superior, which enabled to visualize
the conducted processing steps immediately.
3.2 | Skeletonization
Since there exists no established three‐dimensional skeletonization
method for filamentous microorganisms, we adopted and adapted a
procedure that has been already successfully applied for two‐dimensional image analysis. Barry, Williams, and Chan (2015)
published an ImageJ plugin for the morphological analysis of light
microscope images of filamentous microorganisms. In their two‐dimensional skeletonization procedure, they also applied an iterative
thinning method, as we chose for our three‐dimensional skeletoniza-
tion. An advantage of the usage of an iterative thinning algorithm is
its computational efficiency. Skeletonization alternatives, such as
methods based on Voronoi Covariance Measurement (Grélard,
Baldacci, Vialard, & Domenger, 2017) would result in significantly
higher computation times. However, computational efficiency is
crucial for the analysis of the high data volume of the µCT
measurements. In the present work, skeletonization was achieved
using the plugin “Skeletonize (2D/3D)” for FIJI/ImageJ (Schneider,
Rasband, & Eliceiri, 2012; Schindelin et al., 2012). This plugin is based
on the three‐dimensional thinning algorithm by Lee, Kashyap, and
Chu (1994) and Homann (2007). The basic procedure is to erode the
object’s surface iteratively until only the centerline remains. To
analyze the obtained skeleton voxels of the three‐dimensional image,
the FIJI plugin “Analyze Skeleton (2D/3D)” (Arganda‐Carreras,Fernández‐González, Muñoz‐Barrutia, & Ortiz‐De‐Solorzano, 2010)was used. Depending on their direct neighbors, the skeleton voxels
can be categorized into three different classes: tips, “normal” hyphae,
and branch voxels. The tips are endpoints of the skeleton with only
one neighbor. “Normal” hyphae are skeleton voxels with two
neighbors, and branch voxels are skeleton voxels with at least three
neighbors. As illustrated in Figure 3c,d, the plugin labels each voxel of
the skeleton as either a hyphal voxel (red), a branch voxel (green), or
a tip (yellow). In addition, the plugin calculates the total hyphal length
of the pellet based on the number and location of skeleton voxels.
3.3 | Postprocessing
To correct minor issues arising because of preprocessing and
skeletonization, three postprocessing steps were applied to the
skeleton using MATLAB.
The first issue concerns incorrect short junctions, appearing due
to surface roughness (Figure 4a,b). The deletion of such short
junctions is a commonly described issue when working with skeletons
(e.g. Barry, Chan, & Williams, 2009; Grélard et al., 2017). To delete
F IGURE 2 µCT image of a P.chrysogenum pellet; two‐dimensional
projections of three‐dimensional image;slices (100 µm) of different orientations;center of the slices is the pellet center;image processed with basic rendering only.
such junctions, geodesic distance transform (MATLAB function
“bwdistgeodesic”) was used, which can measure the distance
between a pair of voxels obtained by traversing only the foreground
voxels. Here, the distance was measured as quasi‐Euclidean,
approximating the actual distance between each pair of the
neighboring voxels on the path. First, this procedure ascertained
the distance ,xmin branch to the closest branch voxel for each hyphal
(skeleton) voxel. Afterward, the procedure was used to find the
F IGURE 3 Image processing steps for a
small region of the P. chrysogenum pellet;images are rendered with VGSTUDIOMAX. (a) Raw image: CT data with a low
threshold for gray values. (b) Preprocessedimage: for binarization, threshold for grayvalues is applied; additionally, small
connected objects are deleted. (c)Skeletonized image. (d) Analyzed image:analysis of the skeleton; tips are markedyellow and branches green [Color figure
can be viewed at wileyonlinelibrary.com]
F IGURE 4 Postprocessing steps forsmall regions of P. chrysogenum pellet;transparent white objects illustrate thebinarized three‐dimensional data, red
objects the skeleton, yellow markers thetips, and green markers the branches. (a)Without postprocessing: coarse surface of
binarized hyphae results in short junction.(b) Postprocessing: short junction resultingfrom coarse surface of binarized hyphae is
deleted. (c) Without postprocessing:overlapping hyphae result in branch. (d)Postprocessing: incorrect branch is
deleted. (e) Without postprocessing: closeparallel hyphae can cause bridges imitatingbranches. (f) Postprocessing: incorrectbranches are deleted [Color figure can be
viewed at wileyonlinelibrary.com]
SCHMIDEDER ET AL. | 5
26
distance ,xmin tip to the closest tip voxel. After this,
= +, + , ,x x xsum tip branch min tip min branch for each voxel of the skeleton
was calculated. Tip and “normal” hyphae voxels of the skeleton with
≤, +x xsum tip branch threshold were deleted. Branch voxels of the skeleton
with ≤, +x xsum tip branch threshold were changed to “normal” hyphae
voxels. With 4 µm, xthreshold was set to a value close to the hyphal
In addition, overlapping hyphae misleadingly resulted in branches.
This could be easily corrected because overlapping hyphae result in
four hyphal tubes, whereas a true branching hypha generates three
hyphal tubes connected to the branch (Figure 4c,d). Hence, the
number of hyphal tubes connected to the branch were counted, and a
branch was defined as correct, only if it was connected to three
hyphal tubes.
Binarization of close parallel hyphae may occasionally be
misinterpreted as branches because of small bridges between the
hyphae. The connection between two parallel hyphae can be caused
by two effects: the fusion between hyphae (Read & Lichius, 2009) or
misinterpreting noisy data by image analysis. The investigation of the
actual origin could be the focus of future studies. In both cases, the
branches are not caused by branching itself. Hence, the last
postprocessing step changed the incorrect branches to “normal”
hyphae (Figure 4e,f). Branches closer to another branch at a distance
of less than 6.5 µm (P. chrysogenum) or 5.5 µm (A. niger) were defined
as incorrect. These thresholds were chosen to be 2 µm higher than
the average hyphal diameters (Nielsen, 1993; Packer et al., 1992).
Overall, this processing step prevented an overestimation of the
number of branches.
3.4 | Determination of the local hyphal fraction
The newly defined morphological property, “local hyphal fraction”
(LHF), is a measure of the local accumulation of hyphal material in
filamentous pellets. The LHF is the ratio between the volume of
hyphal material and the total volume with regard to a specified local
volume. Binarized images of the pellets included hyphal voxels and
voxels corresponding to empty space. The LHF was determined by
applying an appropriate filter to the binarized images. For each voxel,
the filter counted the hyphal voxels in its neighborhood. To obtain an
average of the neighborhood, a cube with an edge length of 61 voxels
was used, with the center of the cube containing the target voxel.
The LHF of a voxel is the number of hyphal voxels of the cube divided
by the total number of voxels of the cube. Thus, a voxel with an LHF
of 0.3 implies that the region of the target voxel has a porosity of
70%, and the remaining 30% of the region is filled with hyphal voxels.
Applying this filter method, the LHF of each pellet voxel was
calculated.
To enable visualization of the LHFs at different points, these
values are shown on spherical shells with varying radius. For each
sphere, the LHFs are shown on 4,800 equally‐distributed points, as
illustrated in Figure 9. The points were obtained by the HEALPix
discretization (Gorski et al., 2005), which provides 4,800 representa-
tive coordinates at the sphere surface. Thus, the value of the LHF of
the closest voxel of the pellet was assigned to the representative
coordinate of the sphere.
4 | RESULTS AND DISCUSSION
In this study, we demonstrate the potential of our µCT measurement
and three‐dimensional image analysis system to investigate the
morphology of filamentous pellets. Exemplarily, two pellets have been
investigated in detail: one A. niger (Figure 1) and one P. chrysogenum
(Figure 2) pellet. A. niger represents the coagulative type for pellet
formation and P. chrysogenum the hyphal element agglomerating type.
The investigation of the pellets is based on final binarized (Figure 3a,b),
skeletonized (Figure 3c,d), and post‐processed (Figure 4) pellets. The
F IGURE 5 Skeletonized region of
P. chrysogenum pellet: white transparentobject illustrates the binarizedthree‐dimensional data, red object the
skeleton, yellow markers the tips, andgreen markers the branches [Color figurecan be viewed at wileyonlinelibrary.com]
6 | SCHMIDEDER ET AL.
27
mentioned operation steps are described and discussed in detail in
Section 3. Figure 5 displays a processed region of the P. chrysogenum
pellet.
4.1 | Global morphology of pellets
The final processed images were used to calculate pellet diameter,
porosity, total hyphal length, average hyphal diameter, total tip
number, total branch number, HGU, and hyphal branch unit (HBU). In
the case of the A. niger pellet, a conical section of the pellet not
affected by the fixation material was analyzed (Figure 1a). To
determine the morphology of the whole pellet, the other part of the
pellet was assumed to exhibit the same properties as the studied
portion. This approach was selected because visual observation of
the A. niger showed spherical symmetry (Figure 1 and see below). As
the P. chrysogenum pellet was not affected by the fixation material, it
was possible to analyze the complete pellet.
The diameter of the pellet dPellet was estimated as the volume
equivalent diameter of the convex hull of the respective pellet:
π=
⋅d VPellet
6 ConvHull3 . The porosity, ϵ , was defined by the volume of the
hyphal material VHyphae and the volume of the convex hull of the
pellet: ϵ = −1V
VHyphae
ConvHull. The total hyphal length LHyphae, total tip
number, and total branch number were determined as described
earlier. Based on the volume of the hyphal material and the total
F IGURE 6 Regions with high hyphal
fractions of the cone of the A. niger pelletillustrated for three different perspectives;white: binarized cone of A. niger pellet;
yellow: regions with a hyphal fractionhigher than 0.20. (a) Bottom perspective.(b) Right perspective. (c) Front perspective
[Color figure can be viewed atwileyonlinelibrary.com]
F IGURE 7 Regions with high hyphal
fractions of P. chrysogenum pelletillustrated for three different perspectives;white: binarized P. chrysogenum pellet;
yellow: regions with a hyphal fractionhigher than 0.20. (a) Bottom perspective.(b) Right perspective. (c) Front perspective[Color figure can be viewed at
wileyonlinelibrary.com]
SCHMIDEDER ET AL. | 7
28
hyphal length, the average diameter of hyphae dHyphae was calculated
asπ
=⋅
⋅d V
LHyphae4 Hyphae
Hyphae. The HGU and HBU were estimated as
follows:
=HGUTotal hyphal length
Total number of tips(1)
=HBUTotal hyphal length
Total number of branches. (2)
The morphological properties of the analyzed A. niger pellet and
P. chrysogenum pellet are listed in Table 1. The A. niger pellet has a
diameter of 633 µm, a porosity of 0.87, and an average hyphal
diameter of 3.8 µm. The P. chrysogenum pellet has a diameter of
1085 µm, a porosity of 0.94, and an average diameter of 4.1 µm.
These diameters of the hyphae correspond to those described in
the literature (Colin, Baigorí & Pera, 2013; Morrison & Righelato,
1974; Nielsen, 1993; Packer et al., 1992). Since the average hyphal
diameter was calculated on the basis of the volume of the hyphal
material and the total hyphal length, our image data suggest that
they are sufficiently precise to deduce biologically meaningful
information. The HGU and HBU, respectively, were 95 µm and 93
µm for A. niger and 150 µm and 135 µm for P. chrysogenum.
Considering a single pellet, the HGU and HBU should be in a similar
range because each new branch results in one new tip. With 95 µm,
the A. niger pellet showed an HGU in a range comparable to literature
(Colin et al., 2013). By contrast, the P. chrysogenum pellet, with an
HGU of 150 µm, showed a higher HGU value than mentioned in the
literature (29–99 µm; Morrison & Righelato, 1974). The reported
value, however, has been obtained for disperse mycelia and loose
clumps and therefore cannot be directly compared with the present
study. Nevertheless, we emphasize that future studies involving µCT
measurements of fungal pellets need a large sample size for analysis
to obtain statistically significant data.
It is clear that µCT analysis is a powerful tool to estimate
morphological properties of fungal pellets, including total hyphal
length, total tip number, total branch number, HGU, and HBU. To our
knowledge, this is the first study to do so. Moreover, properties such
as porosity and average hyphal diameter can be estimated on the
basis of information obtained for a complete pellet. The determina-
tion of pellet’s porosity via light microscopic analysis of pellet cross
sections (Hille et al., 2005; Lin et al., 2010) was to date possible only
with low accuracy due to the thickness of the applied slices. Applying
F IGURE 8 Morphological properties of spherically symmetric A. niger pellet dependent on radius; morphological properties are calculatedfor shells. The width of shells is 25 µm and the inner sphere has a radius of 50 µm. (a) Hyphal fraction. (b) Number of tips/branches per volume.
(c) Hyphal growth unit (HGU): hyphal length in micrometer per number of tips; hyphal branch unit (HBU): hyphal length in micrometer pernumber of branches [Color figure can be viewed at wileyonlinelibrary.com]
F IGURE 9 Local hyphal fraction of P.chrysogenum pellet for different distancesfrom the pellet center: (a) 150 µm. (b)
250 µm. (c) 350 µm [Color figure can beviewed at wileyonlinelibrary.com]
TABLE 1 Morphological properties of the fungal pellets
A. niger P. chrysogenum unit
Diameter of pellet 633 1,085 µm
Porosity 0.87 0.94 –
Total hyphal length 1,472,998 2 995,557 µm
Average diameter of hyphae 3.8 4.1 µm
Total number of tips 15,425 20,000 –
Total number of branches 15,768 22,123 –
Hyphal growth unit 95 150 µm
Hyphal branch unit 93 135 µm
8 | SCHMIDEDER ET AL.
29
Gutiérrez‐Correa, 2010) or flow cytometry (Ehgartner et al., 2017),
only superficial parts of pellets have been investigated.
4.2 | Distribution of hyphal material within pelletstructures
Next, we checked whether the three‐dimensional distribution of fungal
biomass within a pellet can be used to deduce information about the
pellet’s genesis and development during submerged cultivation. To
determine the regions with a high biomass content, three‐dimensional
distribution of the LHF was calculated, as described in Section 3.4.
Subsequently, a threshold value for the LHF was set. As a result, only
the regions of the pellet with a higher LHF than the set threshold
remained.
The threshold for the hyphal fraction of the conical section of the
A. niger pellet was set to 0.20 (Figure 6), which was chosen to be close
to the maximum hyphal fraction. Two regions with high hyphal
fractions were observed: a smaller region in the pellet’s center and a
larger region in the outer part. The regions with high hyphal fractions
show spherical symmetry with the cone tip at the center, which was
reflected in Figure 1, confirming that the A. niger pellet was
spherically symmetric. The dense region in the pellet’s center was
possibly a result of agglomeration events involving the conidia and
germ tubes. These early agglomeration events are common pellet‐building mechanisms seen in filamentous microorganisms of the
coagulative type, including A. niger Zhang & Zhang, 2016.
With 0.20 as the threshold of the LHF of the P. chrysogenum pellet,
the value was also set close to the maximum LHF. Figure 7 shows three
regions with high hyphal fractions: one large region and two small ones.
Although the P. chrysogenum pellet on its own is spherical, no spherical
symmetry was visible for the location of the regions with high hyphal
fractions. Considering the three regions with high hyphal fractions, it
can be supposed that the P. chrysogenum pellet was a product of the
agglomeration of three hyphal clumps. This is a typical pellet‐formation
mechanism for filamentous microorganisms of the hyphal element
agglomerating type, including P. chrysogenum (Nielsen, Johansen,
Jacobsen, Krabben, & Villadsen, 1995; Veiter et al., 2018).
Our data show that analyzing hyphal fractions provides new
insights into the symmetry of fungal pellets and hypotheses can be
deduced with respect to their evolution. This knowledge can be used
to rationally engineer aggregation events by appropriate process
control and/or by genetic modifications to obtain final pellet
structures with improved productivities during industrial processes.
4.3 | Radial morphology of spherically symmetricpellets
Spherically symmetric pellets offer the opportunity to characterize
morphological properties along their radii. Thus, the morphological
properties of shells of the A. niger pellet were analyzed. Again, only a
cone of the pellet not affected by the fixation material was studied
(compare Figure 1a). To determine the morphology of the entire
pellet, the remainder part of the pellet was assumed to exhibit
properties identical to those of the cone. The width of shells was set
to 25 µm. The inner sphere had a radius of 50 µm. To determine the
hyphal fraction of a shell, the hyphal volume of the shell was divided
by the total volume of the shell. In addition, the number of branches
and tips per shell were divided by the total volume of shells. To
obtain the HGU and HBU, the hyphal length of the shells was divided
by the number of tips and branches.
Figure 8a shows that the hyphal fraction of the A. niger pellet
contained two maxima. One maximum was in the pellet center, and
the other one at a distance 200 µm outside the center. These
observations are in accordance with the investigation of the regions
with a high hyphal fraction in Section 4.2. Between the two maxima, a
local minimum with a hyphal fraction of about 0.15 could be
observed at 100–125 µm. Starting from the second local maximum at
200 µm, the hyphal fraction decreased, until the edge of the pellet
was reached at a radius between 300 and 325 µm.
With ⋅2.8 105 branches per mm3, the branching density was
highest at the pellet center. From there, the branching density
gradually decreased to ⋅1.9 105 branches per mm3 at a radius of
250 µm. Then, the branching density decreased rapidly toward the
edge of the pellet. With ⋅2.0 105 tips per mm3, the tip density had a
local minimum at the pellet center. From there, the tip density
increased until the maximum tip density of ⋅4.2 105 tips per mm3 was
reached between 100 and 125 µm. From there, the tip density
decreased towards the edge of the pellet.
HGU and HBU result from the hyphal length of shells and the
corresponding number of tips and branches. It is clear that HBU does
not vary strongly along the radius and has values between 80 and
120 µm. Therefore, it can be assumed that the number of branches is
directly proportional to the hyphal length for the whole pellet. By
contrast, the HGU varies strongly along the radius, which is why the
number of tips is not directly proportional to the hyphal length.
4.4 | Nonsymmetry of pellets
Finally, a method to detect and illustrate nonsymmetry of filamentous
pellets is presented. Exemplarily, the method was applied to the P.
chrysogenum pellet. Thereby, the LHF was calculated as described in
Section 3.4. The mass center of the pellet was calculated based on the
hyphal volume elements. By using the HEALPIX algorithm, the LHF for
different distances from the pellet center was determined. Figure 9
shows the LHFs for three different distances from the center.
As shown, the LHF varied strongly for all illustrated distances from
the pellet center. At a distance of 150 µm from the center, the LHF
varied between 0.075 and 0.15. At a distance of 250 µm, the range of
the LHF was between 0.075 and 0.3. At a distance of 350µm, the LHF
varied even more strongly, with values between 0 and 0.3.
4.5 | Critical evaluation of the µCT‐based method
The new method to determine morphological properties with µCT
measurements and subsequent three‐dimensional image analysis has
considerable benefits compared with existing methods. For the first
SCHMIDEDER ET AL. | 9
30
time, hyphal material, tips, and branches can be located and quantified
for entire filamentous pellets. Thus, it is also the first time, that the
following properties of entire pellets are reported: total hyphal length,
total tip number, total branch number, HGU, and the newly defined
parameter HBU.With the applied technique, the inner part of the fungal
pellet can be studied along with the outer part, which is nearly
impossible with light microscopy. This fact makes it difficult to validate
the here newly introduced method with existing methods. However,
applying light microscopy, the influence of the freeze‐drying process hasbeen evaluated. As shown in the Figures S1–S3, the freeze‐drying does
not appear to significantly modify either the size of the pellets or hyphal
thickness. A limitation of the new method is the high costs of µCT
measurements. In addition, three‐dimensional image analysis to obtain
morphological properties follows an elaborate workflow that requires
special computational expertize. As any other image analysis technique,
the here‐described three‐dimensional one has potential sources of
errors/inaccuracies, such as the selection of the gray value threshold
and the selection of the two parameters of the postprocessing steps.
Compared with other methods, such as microscopy and flow cytometry,
the sample number is low and the measurement rather time‐consuming.
µCTs have also lower voxel resolutions compared with pixels of
microscopic approaches.
5 | CONCLUSIONS
The present study describes a new method to determine the
morphological properties of the pellets of filamentous fungi. Freeze‐dried fungal pellets were analyzed using µCT measurements and three‐dimensional image analysis. µCT produces three‐dimensional images,
which is why the inner part of the fungal pellet can be studied along
with the outer part. To our knowledge, this is the first time that total
hyphal length, total tip number, total branch number, and HGU of entire
pellets can be determined with a single technique. Morphological
properties, such as porosity and average diameter of the hyphae, can be
studied in higher detail than ever before by using our method. By
elucidating the spatial morphological distribution, multiple hypotheses
regarding the morphological development of fungal pellets can be
formulated. Summarizing, compared with established methods, the new
method has significant advantages in analyzing the morphology of entire
pellets, including the hyphal network with the location of hyphal
material, tips, and branches. Main disadvantages are the low sample
number and high costs of the new method.
ACKNOWLEDGMENTS
The authors thank the Deutsche Forschungsgemeinschaft (DFG)
for the financial support for this work (BR 2035/11‐1 and ME
2041/5‐1) within the SPP 1934 DiSPBiotech. This work made use
of equipment that was funded by the Deutsche Forschungsge-
meinschaft (DFG INST 95/1111‐1). We also wish to thank Andrea
Pape for assistance with the preparation of pellets and Johann
Landauer for assistance with the microscope images. Strain
P. chrysogenum MUM 17.85 was provided by Prof. Dr. Armando
Venancio, Micoteca da Universidade do Minho, Braga, Portugal.
CONFLICT OF INTERESTS
The authors declare that they have no conflict of interests.
ORCID
Stefan Schmideder http://orcid.org/0000-0003-4328-9724
REFERENCES
Arganda‐Carreras, I., Fernández‐González, R., Muñoz‐Barrutia, A., &
Ortiz‐De‐Solorzano, C. (2010). 3D reconstruction of histological
sections: Application to mammary gland tissue. Microscopy Research
and Technique, 73, 1019–1029. https://doi.org/10.1002/jemt.v73:11
Barry, D. J., Chan, C., & Williams, G. A. (2009). Morphological quantification
of filamentous fungal development using membrane immobilization and
automatic image analysis. Journal of Industrial Microbiology & Biotechnol-
ogy, 36, 787. https://doi.org/10.1007/s10295‐009‐0552‐9Barry, D. J., Williams, G. A., & Chan, C. (2015). Automated analysis of
filamentous microbial morphology with AnaMorf. Biotechnology
Salvo, L., Cloetens, P., Maire, E., Zabler, S., Blandin, J. J., Buffière, J. ‐Y., &Josserond, C. (2003). X‐ray micro‐tomography an attractive character-
isation technique in materials science. Nuclear Instruments and Methods in
Physics Research B: Beam Interactions with Materials and Atoms, 200, 273–
286. https://doi.org/10.1016/S0168‐583X(02)01689‐0Schindelin, J., Arganda‐Carreras, I., Frise, E., Kaynig, V., Longair, M.,
Pietzsch, T., & Cardona, A. (2012). Fiji: An open‐source platform for
Yang, H., King, R., Reichl, U., & Gilles, E. D. (1992). Mathematical model for
apical growth, septation, and branching of mycelial microorganisms.
Biotechnology and Bioengineering, 39, 49–58. https://doi.org/10.1002/
bit.260390109
Zhang, J., & Zhang, J. (2016). The filamentous fungal pellet and forces
driving its formation. Critical Reviews in Biotechnology, 36, 1066–1077.
https://doi.org/10.3109/07388551.2015.1084262
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Supporting Information section at the end of the article.
How to cite this article: Schmideder S, BarthelL, Friedrich T,
et al. An X‐ray microtomography‐based method for detailed
analysis of the three‐dimensional morphology of fungal
pellets. Biotechnology and Bioengineering. 2019;1–11.
https://doi.org/10.1002/bit.26956
SCHMIDEDER ET AL. | 11
32
Supplementary Materials for Paper I: AnX-ray microtomography-based method fordetailed analysis of the three-dimensionalmorphology of fungal pellets
Figure S1: Light microscope images of A. niger pellets after 48 h fermentation: theupper row shows the “wet” states of pellets, whereas the lower row shows the freeze-dried states; Columns contain microscope images of the same pellet for the two states“wet” and freeze-dried.
33
Figure S2: Light microscope images of P. chrysogenum pellets after 48 h fermentation:the upper row shows the “wet” states of pellets, whereas the lower row shows the freeze-dried states; Columns contain microscope images of the same pellet for the two states“wet” and freeze-dried.
Figure S3: Light microscope images of P. chrysogenum pellets after 24 h fermentation:the upper row shows the “wet” states of pellets, whereas the lower row shows the freeze-dried states; Columns contain microscope images of the same pellet for the two states“wet” and freeze-dried.
34
Figure S4: Light microscope images of a P. chrysogenum pellet applying high resolutions:the upper row shows the “wet” state of the pellet, whereas the lower row shows thefreeze-dried state.
Figure S5: Light microscope images of outer regions of different P. chrysogenum pelletsapplying high resolutions: the upper row shows the “wet” state of pellets, whereas thelower row shows the freeze-dried state.
35
5.2 Paper II: From three-dimensional morphology to effective
diffusivity in filamentous fungal pellets (Schmideder et al.,
2019b)
Summary
The supply of pellets with nutrients and oxygen strongly influences their growth and pro-
ductivity. While the consumption of nutrients and oxygen is required for fungal metabolism,
their transport into pellets is mainly driven by diffusion. Although the dense hyphal network
is known to limit the diffusive mass transport of nutrients, oxygen, and secreted metabolites,
a well-founded correlation between structure and diffusivity does not exist. In this study,
we computed the effective diffusivities through a few hundred representative sub-volumes
of five Aspergillus niger pellets. The three-dimensional structures of the pellets were deter-
mined with µCT measurements and image analysis described in Paper I. Based on the diffu-
sion computations, we obtained a correlation between effective diffusivity and solid hyphal
fraction. This correlation is inspired by material laws for fibers, consistent with theoreti-
cal expectations, and shows an excellent fit to the investigated A. niger pellets. While the
correlation uncovered discrepancies with previously assumed laws for filamentous fungi, it
showed some similarities to laws for randomly oriented fibers. The findings of this study
enable the prediction of the diffusive transport of nutrients, oxygen, and secreted metabolites
in filamentous fungal pellets. This knowledge will improve morphological engineering of
pellets, and thus, contribute to increased productivities in bioprocesses. Restrictions of the
applied µCT system demanded the application of a fixed resolution for the measurements.
Further, pellets originated from a single experimental setup, which resulted in comparable
pellet-micromorphologies. However, both resolution and micromorphology can alter the dif-
fusivity. These limitations were overcome in Paper III.
Author contributions
Stefan Schmideder did the conception and design of the study and wrote the manuscript,
which was edited and approved by all authors. Heiko Briesen and Vera Meyer supervised the
study. Lars Barthel cultivated and freeze-dried fungal pellets. Stefan Schmideder and Henri
Müller conducted µCT measurements and image analysis. Stefan Schmideder set up mor-
phological simulations and diffusion computations. Stefan Schmideder and Heiko Briesen
interpreted the results. Stefan Schmideder and Lars Barthel proposed a workflow for biopro-
any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.
37
other hand, the production of secondary metabolites peaks when the
producing organism shows an extremely low or zero growth (Brakhage,
2013). These conditions can be observed for example in the dense center
of fungal pellets, where the restricted diffusion of oxygen and nutrients
leads to limitations, thus inhibiting growth (Veiter, Rajamanickam, &
Herwig, 2018). This demonstrates that a detailed understanding of the
limitations and diffusion processes in fungal pellets is of crucial
importance for many biotechnological applications.
The effective diffusion coefficient is required to calculate the
diffusive mass transport. For component i in a porous medium, this
parameter can be expressed as (Becker, Wieser, Fell, & Steiner, 2011)
= ⋅D D k ,i i,eff ,bulk eff (1)
where Di,bulk is the diffusion coefficient of component i in the bulk
medium without geometrical hindrance and keff describes the reduction
of the free bulk diffusion Di, bulk to the effective diffusion
Di, eff and is solely dependent on the pore geometry and independent
of the diffusing substance i (Becker et al., 2011). Hereafter, keff is called
the effective diffusion factor. In the case of diffusion in filamentous
pellets, Di, bulk corresponds to the molecular diffusion coefficient of
components such as oxygen, glucose, or products in the fermentation
medium and is strongly dependent on the medium, diffusing substance,
and process conditions such as temperature. Temperature‐dependentbulk diffusion coefficients for glucose, oxygen, and several other
compounds in aqueous solutions can be estimated, for example, from
Yaws (2014). The geometrically caused reduction of the diffusion, keff,
can be expressed as (Epstein, 1989):
τ=ϵ
k ,eff 2(2)
where the porosity, ϵ, is defined as the ratio of volume of voids to the
total volume. The tortuosity, τ , is a geometrical parameter and can be
taken as the ratio of the average pore length to the length of the
porous medium along the major flow or diffusion axis. Thus, in
general it is > 1 (Epstein, 1989).
So far, no experimental or modeling approaches to exactly predict the
diffusive mass transport inside whole fungal pellets are available. Several
modeling approaches of filamentous microorganisms consider the
diffusive mass transport of oxygen and/or substrates like
glucose (Buschulte, 1992; Celler, Picioreanu, van Loosdrecht, & van
Wezel, 2012; Cui, Van der Lans, & Luyben, 1998; Lejeune &
Baron, 1997; Meyerhoff, Tiller, & Bellgardt, 1995; Table 1). They include
effective diffusion coefficients that are dependent on the molecular
TABLE 1 Applied correlations between effective diffusion coefficients (Di,eff), effective diffusion factor (keff), bulk diffusion coefficients(Di,bulk), porosity (ϵ), hyphal fraction ( = − ϵc 1h ), solid fraction (ϕ), and tortuosity (τ ) in the literature to model the diffusive transport
mechanisms
Eq. for keff in =D D ki i,eff ,bulk eff
Originstructure Diffusion direction Field Source
− c1 h – – Filamentous microorganisms Celler et al. (2012)
– – Filamentous microorganisms Cui et al. (1998) adopted from Van’t
Riet and Tramper (1991)
– – Filamentous microorganisms Buschulte (1992) adopted from Aris
(1975)
τ
( − )c1 h with τ = 2 – – Filamentous microorganisms Lejeune and Baron (1997)
−
+
c
c
2 2
2h
h
– – Filamentous microorganisms Meyerhoff et al. (1995) adopted from
Neale and Nader (1973)
(− )cexp 2.8 h – – Filamentous microorganisms Buschulte (1992) adopted from Vorlop
(1984)
ϕ
ϕ ϕϕ
ϕ
⎛
⎝
⎜⎜−
⎞
⎠
⎟⎟+ − −
−
1 2
1 0.013360.30583 4
1 1.40296 88
– Perpendicular to
fibers
Square array of parallel fibers Perrins et al. (1979)
Baumgartl, Lübbers, & Schügerl, 1986). The researchers correlated the
mass transport of oxygen into the pellets with microscopic information
from cryo‐slices. However, the microscopy images of the cryo‐slicesmissed the three‐dimensional information of the hyphal network and
hyphae superimposed in the two‐dimensional projections. Therefore,
the appropriate correlation between pellet morphology and diffusion was
not possible in this experimental approach. We could recently overcome
the limitations of two‐dimensional image generation by performing X‐raymicrocomputed tomography (μCT) measurements on whole fungal pellets
(Schmideder et al., 2019). The investigated Aspergillus niger pellet showed
a very complex three‐dimensional structure. With a diameter of μ633 m,
the A. niger pellet had a hyphal length of 1.5m and 15,425 tips in total.
In materials science, μCT measurements and subsequent
mass‐transfer computations matured into a widely used method
to determine the effective diffusivity of porous/fibrous materials.
Exemplarily, Panerai et al. (2017), Becker et al. (2011), Coin-
dreau, Mulat, Germain, Lachaud, and Vignoles (2011), and
Foerst et al. (2019) investigated the effective diffusivity of
fibrous insulators, fuel cell media, carbon‐carbon composites, and
maltodextrin solutions, respectively. Thereby, μCT appeared as a
non‐destructive technique to measure three‐dimensional micro‐structures.
In this study, we computed the effective diffusivity of A. niger
pellets based on the micro‐structural characterization gained from
μCT measurements and subsequent image analysis. The newly
developed technique for fungal pellets combines the experimental
acquisition of three‐dimensional images with the locally resolved
calculation of the effective diffusion coefficient and tortuosity within
this structure for the first time. This results in an unprecedented
potential for the determination of diffusion processes inside fungal
pellets.
2 | MATERIALS AND METHODS
2.1 | Preparation of pellets
The A. niger hyperbranching strain MF22.4, which has been shown to
be a better protein‐secretion strain than the wild‐type strain (due to
deletion of the racA gene; Fiedler, Barthel, Kubisch, Nai, & Meyer,
2018), was used in this study. Pellets were obtained by submerged
cultivation of MF22.4 for 48 hr and freeze‐drying following the
previously described protocol (Schmideder et al., 2019).
2.2 | X‐ray microcomputed tomography
To obtain three‐dimensional images of the freeze‐dried filamentous
pellets, μCT measurements were performed based on the method
reported by Schmideder et al. (2019). Two‐dimensional projections
from different angles were reconstructed to generate three‐dimen-
sional images with a custom‐designed software (Matrix Technologies,
Feldkirchen, Germany) that uses CERA (Siemens, Munich, Germany).
The image resolution was 1 μm (i.e., the edge length of the voxels was
1 μm), and to generate the beam, 60 kV and 25 μA were applied.
Depending on the size of the pellets, 1‐5 pellets can be measured
with one μCT‐measurement (3 hr including the time for image
reconstruction). An instant adhesive (UHU, Bühl, Germany) was used
to fix the freeze‐dried fungal pellets on top of a sample holder. In
contrast to the previous study (Schmideder et al., 2019), in this case,
the instant adhesive dried 5min before placing the pellets on top of
the holder. This procedure resulted in a smooth surface of the instant
adhesive, while it remained sticky enough to fix the pellets. The
smooth surface facilitated the segmentation of the instant adhesive
in the subsequent image processing.
2.3 | Image processing
Image processing aimed at differentiating between hyphal material
and background. The background included the instant adhesive used
for sample fixation, the air between the hyphae, and small impurities.
In general, image processing is leaned to the one reported in the
section “Preprocessing” by Schmideder et al. (2019)). The image
processing result for one pellet is exemplarily illustrated in Figure 1.
As the instant adhesive showed similar gray values as the pellets,
we did not implement an automated segmentation. Instead, the
instant adhesive at the bottom of the pellets was cropped manually
using the commercial software VGSTUDIO MAX (version 3.2,
Volume Graphics GmbH, Heidelberg, Germany) in a first processing
step. The further image processing steps were carried out auto-
matically using MATLAB (version 2018a, MathWorks, Natick, MA).
To differentiate between hyphae and air voxels, a threshold –
calculated by Otsu’s method (Otsu, 1979) – was applied on the gray
value images. Finally, small connected objects with a maximum size of
1,000 μm3 were deleted to eliminate objects that were not part of
the pellet. The processed three‐dimensional pellets were used for
further diffusion computations.
SCHMIDEDER ET AL. | 3
39
2.4 | Representative cubes for diffusioncomputations of A. niger pellets
To compute the spatially resolved effective diffusivity in fungal pellets,
we extracted representative cubic sub‐volumes of the processed
three‐dimensional images. The diffusive mass transport was computed,
as described in Section 2.6 using these cubes. The centers of the cubes
used for diffusion computations were selected along the main axis
originating from the calculated mass center of the A. niger pellets. The
distance between the centers of the cubes along the main axis was set
to 25 μm. In Figure 2, the distance between the cubes was increased
for the sake of clarity. To identify the influence of the cube‐sizeon the diffusion computations, the edge length of the cubes was varied
to 30, 50, 70, and 90 μm.
2.5 | Beam‐Pellet
A common assumption for the effective diffusion coefficient
of filamentous microorganisms is the direct proportionality
to the porosity ϵ (Buschulte, 1992; Celler et al., 2012; Cui
et al., 1998; Silva, Gutierrez, Dendooven, Hugo, & Ochoa‐Tapia,2001):
= ⋅ ϵD D .i i,eff ,bulk (3)
Thereby, the effective diffusion factor keff is assumed to be equal to
the porosity, and the tortuosity is neglected. To imitate a filamentous
spherical object, where the tortuosity can be nearly neglected, we
simulated a so‐called “Beam‐Pellet,” which was used to validate the
F IGURE 1 Processed three‐dimensional μCT image of an Aspergillus niger pellet. The images were rendered using VGSTUDIO MAX.
(a) Projection of the whole pellet. (b) Projection of a central slice with a depth of 25 μm [Color figure can be viewed at wileyonlinelibrary.com]
F IGURE 2 Processed three‐dimensional μCT image of the Aspergillus niger pellet of Figure 1 and cubes for the diffusion computations:(a) Transparent: projection of a whole pellet; red: exemplary cubes that were used for the diffusion computations. (b–e) Morphology of asingle cube from different viewing directions; the gray boundaries in (c–e) illustrate the boundaries parallel to the diffusion computation.
(b) Cube without illustration of boundaries for the diffusion computations. (c) Cube for diffusion in the x‐direction. (d) Cube for diffusion in they‐direction. (e) Cube for diffusion in the z‐direction [Color figure can be viewed at wileyonlinelibrary.com]
4 | SCHMIDEDER ET AL.
40
method of the diffusion computation and critically scrutinize the
tortuosity neglect in the literature.
The “Beam‐Pellet” (Figure 3) was built up from equally sized
filaments (in diameter and length) with their origin in the pellet
center and having a radial orientation. To guarantee a uniform
distribution of the filaments in space, their orientation was calculated
using the HEALPIx (hierarchical equal area iso‐latitude pixelization)
discretization (Gorski et al., 2005). In this way, 50,700 representa-
tive, equally distributed points were calculated on the pellet surface;
all of them were connected to the pellet center. Then, the connected
lines were dilated with MATLAB to obtain a “Beam‐Pellet” with a
defined diameter of the filaments. The diameter was chosen to be
3 μm, similar to the average hyphal diameter of A. niger (Colin,
Baigorí, & Pera, 2013; Nielsen, 1993; Schmideder et al., 2019). To
investigate the influence of the image resolution on the subsequent
diffusion computations, similar “Beam‐Pellets” with different resolu-
tions were simulated. Starting from a ”Beam‐Pellet” with a radius of
700 voxels and a dilation of one voxel, the resolution of the filaments
was increased. Thereby, the number of filaments was kept constant,
whereas the radius of the “Beam‐Pellet” was set to 1,167, 1,633,
2,567, and 3,500 voxels and the dilation was set to 2, 3, 5, and
7 voxels, respectively. Thus, the “Beam‐Pellets” only differed in the
scaling and the resolution of the filaments. The effective diffusivity of
the “Beam‐Pellet” was analyzed, as described in Section 2.6. This
analysis required cubes that represent the whole pellet. Similar to the
case of the A. niger pellets, cubes located along the main axes were
chosen to investigate the effective diffusivity (Figure 3). Identical to
the final analysis of the A. niger pellets, the cube‐edge length was set
to 50 μm, and the cubes were selected along the main axis. Thus, the
cube edge length was 50 voxels for the “Beam‐Pellet” with a radius of
700 voxels and a dilation of one voxel. To guarantee that the same
structure (50 μm× 50 μm× 50 μm) was analyzed for higher
resolutions, we increased the cube edge length to 83, 117, 183,
and 250 voxels, respectively. In Figure 4, the same representative
cube is shown for different resolutions.
2.6 | Computation of the effective diffusivity
To compute the effective diffusion factor and tortuosity of
filamentous structures (Equation (2)), the module DiffuDict of the
commercial software GeoDict (Becker et al., 2011; Velichko,
Math2Market Gmbh, Kaiserslautern, Germany) was used. As
DiffuDict allows for the voxel‐based solution of transport equa-
tions, the processed three‐dimensional image data of the μCT
measurements as well as the simulated “Beam‐Pellet,” could be
used for the diffusion computations. DiffuDict requires cubic
domains for the diffusion computations, and therefore, cubic
sub‐volumes of the filamentous pellets were extracted from the
three‐dimensional images for further analysis. The selections of
representative sub‐volumes are described in Sections 2.4 and
2.5 for A. niger pellets and the “Beam‐Pellet,” respectively.
Conceptually, as shown in Figures 2c–e and 3c–e, the diffusion
computations in DiffuDict were executed in one of the three main
axes for each computation. As the representative cubes were chosen
along the main axes, we were able to apply three computations on
each cube and could thus analyze the effective diffusivity of each
cube in the radial and in two tangential directions. The computa-
tional effort to obtain the diffusivity of one cube in the three
directions was about 1 min for cubes with 50 × 50 × 50 voxels with
an Intel Xeon E5‐1660 CPU (3.7 GHz). The details of the computa-
tions are described in the following.
F IGURE 3 Simulated “Beam‐Pellet” and cubes for diffusion computations: (a) Transparent: projection of the whole pellet; red: exemplarycubes that were used for the diffusion computations. (b–e) Morphology of a single cube from different viewing directions; the gray boundaries
in (c–e) illustrate the boundaries parallel to the diffusion computation. (b) Cube without illustration of boundaries for the diffusioncomputations. (c) Cube for diffusion in the x‐direction. (d) Cube for diffusion in the y‐direction. (e) Cube for diffusion in the z‐direction [Colorfigure can be viewed at wileyonlinelibrary.com]
SCHMIDEDER ET AL. | 5
41
We computed the diffusion in the space/liquid between the
hyphae (porous medium). The predominant diffusion regime in
liquids is bulk diffusion (Becker et al., 2011; Panerai et al., 2017),
that is, mass transport is mainly driven by collisions between fluid
molecules. In our approach, we neglected surface effects on the
solid–liquid interface that could influence diffusion. One possible
effect could be surface diffusion, that is molecules can diffuse on
the surface of pores. This phenomenon is known to be an
important transport mechanism in reversed‐phase liquid chroma-
tography. However, predictions are difficult and depend on the
temperature, surface concentration, and surface chemistry
1993) and thus similar but not identical to our “Beam‐Pellets.”In contrast to the tangential diffusion paths, the radial paths of the
“Beam‐Pellet” were not winding (Figure 3). In theory, increased path
lengths result in an increased tortuosity and a decreased effective
diffusion factor (Epstein, 1989). This behavior could be observed for the
“Beam‐Pellet” in the radial and tangential diffusion directions as well.
The small difference between the computed radial effective diffusion
factors/tortuosities and the “law of mixtures” for the flow along parallel
fibers (Tomadakis & Robertson, 2005) = ϵkeff was evoked by the radial
direction of the filaments of the “Beam‐Pellet.” Thus, the filaments were
not completely parallel to each other. Further investigations with
simulated perfectly parallel filaments resulted in = ϵkeff and a constant
tortuosity of one (data now shown).
To sum up, the diffusion behavior of the “Beam‐Pellet” was
consistent with the diffusion theory described by Epstein (1989) and
could be used to verify the applied diffusion computations. The
comparison to literature correlations about the diffusivity of parallel
fibers suggests that our diffusion computations of fibers with a low
image resolution tend to underestimate the diffusivity, whereas high
resolutions approach the literature correlations. The radial diffusion
behavior of the “Beam‐Pellet” illustrated the often applied literature
assumption for filamentous microorganisms: = ϵkeff (Buschulte, 1992;
Celler et al., 2012; Silva et al., 2001; Van’t Riet & Tramper, 1991).
However, the morphology of the idealized “Beam‐Pellet” (Figure 3) was
very different from the μCT data of A. niger pellets (Figure 2). Thus, the
actual correlation between the effective diffusivity and the hyphal
fraction of A. niger pellets was investigated in further detail.
3.2 | Effective diffusivity of A. niger pellets
Contrary to the “Beam‐Pellet” (Figure 3), in the case of the A. niger
pellets, the voids, that is, the spaces between the hyphae (Figure 2),
(a) (b)
F IGURE 5 Diffusion computations of “Beam‐Pellets” in the radial and tangential directions and comparison to existing literature correlations forparallel fibers (Table 1). The “Beam‐Pellets” differ in the fiber‐diameter in voxels, whereas the diameter in μm stayed constant (Figure 4). Each data pointresults from a single diffusion computation of a cubic sub‐volume with a cube‐edge length of 50 μm. Crosses and circles correspond to computed
diffusion properties in the radial and tangential directions, respectively. The black lines represent literature correlations. The hyphal fraction correspondsto the ratio of the volume of hyphae to the total volume in the cubic sub‐volumes [Color figure can be viewed at wileyonlinelibrary.com]
SCHMIDEDER ET AL. | 7
43
are strongly winded. According to Epstein (1989) that should result
in an increased tortuosity, and therefore, in a decreased effective
diffusivity. The tortuosity and effective diffusion factors of five A.
niger pellets are investigated in this section. The diameters of the
pellets were between 410 and 570 μm.
To investigate the influence of the size of the cubic sub‐volumes
on the diffusion computations, the edge length of the cubes was
varied. Figure 6a shows the relation between the effective diffusion
factor and the hyphal fraction for different cube sizes. The data
include the diffusion computations of the five investigated A. niger
pellets in the radial direction for cube‐edge lengths of 30, 50, 70, and
90 μm. Generally, the effective diffusion factors for different cube‐edge lengths showed similar behavior. When no hyphal material is
present, that is, when the hyphal fraction is zero, the effective
diffusion factor is one, and thus, the diffusion is not geometrically
hindered. With increasing hyphal fraction, the effective diffusion
factor decreases. The applied cube‐edge lengths did not have a high
impact on the diffusion results. Thus, a cube‐edge length of 50 μm
was applied for further computations in this study. Figure 6b) shows
the effective diffusion factors of the five A. niger pellets studied
herein for a cube‐edge length of 50 μm. It can be seen that the values
for the different pellets varied only slightly. As five pellets were
investigated, this very subtle scattering of the data points implies
that the method is quite reproducible when applied to different
pellets obtained from the same cultivation sample.
Figure 7 shows the results of the diffusion computations for the
five studied A. niger pellets in the radial and tangential directions for
a cube‐edge length of 50 μm. Figure 7a shows that the computed
effective diffusion factors of the A. niger pellets were much smaller
than the values expected from the literature assumption = ϵkeff , and
therefore, also much smaller than the values obtained for the
investigated “Beam‐Pellet.” A slight anisotropy was observed when
comparing the radial and tangential diffusion directions because the
effective diffusion factors in the radial direction were slightly higher
(a) (b)
F IGURE 6 Correlations between the radial effective diffusion factor and hyphal fraction (ratio of the volume of hyphae to the total volume) for fiveAspergillus niger pellets. Each data point results from a single diffusion computation of a cubic sub‐volume: (a) Diffusion computations with different cubesizes. (b) The cube‐edge length was 50 μm; each color represents the investigated cubes of one pellet [Color figure can be viewed at
wileyonlinelibrary.com]
(a) (b)
F IGURE 7 Diffusion computations of five Aspergillus niger pellets in the radial and tangential directions. Each data point results from a singlediffusion computation of a cubic sub‐volume with a cube‐edge length of 50 μm; keff is the effective diffusion factor and ϵ is the porosity: (a) The
blue and green data points correspond to effective diffusion factors in the radial and tangential directions, respectively; the black linerepresents the literature assumption = ϵkeff . (b) The blue and green data points represent the tortuosities in the radial and tangential
directions, respectively [Color figure can be viewed at wileyonlinelibrary.com]
8 | SCHMIDEDER ET AL.
44
than their counterparts in the tangential direction. In the absence of
hyphal material, that is, when the hyphal fraction is zero, the
tortuosity is one (Figure 7b). With increasing hyphal fraction, the
tortuosity increases as well. Again, a slight difference was observed
between the radial and tangential diffusion directions, with the radial
diffusion computations resulting in lower tortuosities than their
tangential counterparts. According to Equation (2), the lower
tortuosities explain the higher effective diffusion factors in the radial
direction. In the model assumption = − + = ϵk c 1eff h , the tortuosity
is assumed to be constantly one. This simplification explains
the differences between our computed effective diffusion factors
for the A. niger pellets and the values expected from the literature
(Figure 7a) as well as those reported for the “Beam‐Pellet” in the
previous section.
The local hyphal fraction range of the five investigated A. niger
pellets was 0–0.4. To the best of our knowledge, there are no
reports in the literature, in which the hyphal fraction of filamentous
pellets is higher than the maximum hyphal fraction measured in this
study, for example, Cui, Van der Lans, and Luyben (1997, 1998)
reported average hyphal fractions between 0.07 and 0.30 for whole
Aspergillus awamori pellets. Thus, the hyphal fraction ranges of the
investigated A. niger pellets could already be representative for
realistic pellets.
3.3 | Correlation between effective diffusivity andhyphal fraction
The diffusion computations of the five investigated A. niger pellets
(Section 3.2) are compared to literature assumptions (Table 1) for the
correlation between the effective diffusion factor (keff) and the
microorganisms (Figure 8a) and fibers (Figure 8b). As the diffusion of
spherical pellets should be driven mainly by radial diffusion, we have
investigated it herein. Additionally, a new modeling approach was
introduced to correlate the effective diffusion factor and the hyphal
fraction.
The computed diffusion factors differed strongly from the
assumption = ϵkeff , which has been used for simulation/modeling
studies of filamentous microorganisms by Celler et al. (2012),
Buschulte (1992), and Silva et al. (2001). In fact, the computed
effective diffusion factors were far below the expected values. In
those previous models, the effective diffusion factor was assumed to
be only dependent on the porosity of the material, while neglecting
the tortuosity. Thus, the difference between our computed effective
diffusion factors and previous assumptions (Buschulte, 1992; Celler
et al., 2012; Silva et al., 2001) was not surprising. The second linear
literature assumption for filamentous microorganisms was = ϵ/k 2eff
(Lejeune & Baron, 1997). In their work, Lejeune and Baron (1997)
considered the tortuosity to be consistently 2, without explaining
that assumption. As shown, their model differed strongly from the
computed data. Their model would result in a geometrically hindered
diffusion for a hyphal fraction of zero. Thus, especially for low hyphal
fractions, that assumption seems to be untenable. In their growth
modeling approach for filamentous microorganisms, Meyerhoff et al.
(1995) applied a nonlinear correlation between the effective
diffusion factor and the hyphal fraction: = ( − )/( + )k c c2 2eff h h . This
approach seemed to approximate our computed data better than the
two previous model assumptions from the literature. Additionally,
the effective diffusion factor is 1 and 0 for hyphal fractions of 0 and
1, respectively. In theory, these conditions should be fulfilled.
However, the model seemed to overestimate the effective diffusion
factors. Besides the assumption = ϵkeff , Buschulte (1992) deduced a
second model for filamentous microorganisms: = −k e ceff
2.8 h. In
comparison with the other literature assumptions for filamentous
microorganisms, this approach fitted our computed data best.
However, in the hyphal fraction range of 0–0.4, the model seemed
to underestimate the computed data. Additionally, for a hyphal
fraction of 1, the model would result in an effective diffusion factor of
(a) (b)
F IGURE 8 Correlations between hyphal fraction (ch)/solid fraction/porosity (ϵ = − c1 h) and effective diffusion factor (keff). The blue datapoints correspond to the computed effective diffusion factors of five Aspergillus niger pellets in the radial direction, with the cube‐edge length forthe diffusion calculations being 50 μm. The solid bold blue line shows the new correlation between the hyphal fraction and the effective
diffusion factor; the black lines represent existing correlations in the literature (Table 1) for (a) filamentous microorganisms and (b) 3D randomdistributed overlapping fibers [Color figure can be viewed at wileyonlinelibrary.com]
SCHMIDEDER ET AL. | 9
45
0.06. In theory, for a hyphal fraction of 1, the effective diffusion
factor has to be 0 (Epstein, 1989).
The correlations for 3D random distributed overlapping fibers
(Figure 8b) fitted our data better than the correlations existing for
filamentous microorganisms. Our data fall in between the correlation
of Tomadakis and Sotirchos (1991) and He et al. (2017). In both
studies, the computation of the effective diffusivity was well
validated for other structures like parallel fibers. However, they
differ significantly. Tomadakis and Sotirchos (1991) applied a
modification of Archie’s law (Archie, 1942) to correlate the tortuosity
factor τ α= (( − ϵ )/(ϵ − ϵ ))12p p , with the percolation porosity
ϵ = 0.037p and α = .661. It has to be mentioned that the modification
of Archie’s law also fitted well for Vignoles et al. (2007) for bulk
diffusion of μCT‐generated images of parallel fibrous carbon‐carboncomposite preforms. The percolation porosity was ϵ = 0.04p and
α α= ∕ =0.107 0.465 for diffusion parallel/perpendicular to the
fibers, respectively. According to Nam and Kaviany (2003), the
effective diffusivity of isotropic structures is often estimated using a
power function of porosity. Thus, He et al. (2017) fitted their
diffusion results of 3D random distributed overlapping fibers and
found α= ϵk neff , with α = 1.05 and =n 3.
To overcome the limitations and/or inaccuracies in the correla-
tions between the effective diffusion factor and hyphal fraction for
filamentous microorganisms and to obtain a relation with only one
fitting parameter, we propose a new correlation:
= ( − )k c1 ,aeff h (4)
where a is the only fitting parameter. This rather simple expression
guarantees theory‐consistent effective diffusion factors of 1 and 0 at
hyphal fractions of 0 and 1, respectively, and provides an excellent fit
to our data. Minimizing the error squares, = ±a 2.02 0.02 (with 95%
confidence bounds):
= ( − ) ±k c1 .eff h2.02 0.02 (5)
In conclusion, the literature assumptions for modeling diffusion in
filamentous pellets failed to fit our computed results, whereas
correlations for fibers in material science fitted our data better. Thus,
we set up a nonlinear equation (5) approach with only one fitting
parameter, which is a special case for the power function of porosity, for
α = 1 as well as for the modified Archie’s law, when ϵ = 0p . This
approach modeled the correlation between the effective diffusion factor
and the hyphal fraction for five investigated A. niger pellets quite well.
3.4 | Proposed workflow in bioprocessdevelopment
Depending on the product of interest, a saturation or a limitation
of substrates inside fungal pellets is pursued (Veiter et al., 2018).
To predict the spatial distribution of substrates inside pellets, the
effective diffusivity through the fibrous network has to be known
(Buschulte, 1992; Celler et al., 2012). Thus, our newly proposed
method to determine the effective diffusivity of filamentous fungal
pellets with μCT measurements and subsequent diffusion compu-
tations through the three‐dimensional morphology has consider-
able benefits for bioprocess development. We propose the
following idealized workflow to achieve an optimal/suitable pellet
morphology:
(1) Generate pellets through experiments or simulations
a) Cultivate different strains at different process conditions;
apply μCT measurements and subsequent image analysis of
pellets (Schmideder et al., 2019)
b) Simulate various three‐dimensional pellet‐networks with
algorithms similar to Celler et al. (2012); model calibration
could be carried out based on μCT measurements with
subsequent image analysis (Schmideder et al., 2019)
(2) Compute the correlation between the hyphal fraction and the
effective diffusivity for each existing and simulated pellet of
Step 1, as described in the present study
(3) Compute the proportion of substrate‐limited and substrate‐saturated regions of each pellet based on the consumption‐ anddiffusion‐terms of models such as Buschulte (1992)); apply
correlation of the hyphal fraction and effective diffusivity in
diffusion terms (this study; Step 2)
(4) Assemble data base of experimentally or simulatively generated
pellets including substrate‐supply and morphological features
(5) Select optimal/suitable pellet for the desired process from data base
(6) Realize optimal/suitable pellet in bioprocess, for example,
through the upscale of previous experiments (Step 1a), genetic
modifications, or process control
Obviously, some of these steps have to be investigated and
elaborated in much more detail to reach the proposed optimum
macromorphology through this workflow. However, we consider the
investigation of the effective diffusivity as an important step towards
morphological engineering. In our study, we investigated five pellets
of one process and the correlation between the effective diffusivity
and hyphal fraction scattered only slightly. Thus, we propose, that
our method is at least reproducible for a certain strain at certain
process conditions. If the observed correlation between the hyphal
fraction and the effective diffusivity is representative for the applied
A. niger strain in general, other fungal strains, and/or theoretically all
filamentous microorganisms should be the focus of future studies.
Thereby, as described, μCT measurements are suitable to detect the
three‐dimensional morphology used for diffusion computations.
However, other three‐dimensional methods such as confocal laser
scanning microscopy of pellets slices or even simulated pellets are
also conceivable to explore other strains and processes.
4 | CONCLUSIONS
The findings described in this manuscript unveil the actual relation
between the hyphal fraction (ch, i.e., the ratio between the volume of
10 | SCHMIDEDER ET AL.
46
hyphae and the total volume) and the effective diffusion factor (keff)
and tortuosity inside filamentous fungal pellets. They also uncover a
discrepancy with the assumptions made in the literature for
filamentous microorganisms so far. We propose a new correlation,
which is inspired by correlations for fibers, rather simple, consistent
with theoretical expectations, and shows an excellent fit to the
Schmideder, S., Barthel, L., Friedrich, T., Thalhammer, M., Kovačević, T.,Niessen, L., & Briesen, H. (2019). An x‐ray microtomography‐basedmethod for detailed analysis of the three‐dimensional morphology of
fungal pellets. Biotechnology and Bioengineering, https://doi.org/10.
1002/bit.26956
Silva, E. M. E., Gutierrez, G. F., Dendooven, L., Hugo, J. I., &
Ochoa‐Tapia, J. A. (2001). A method to evaluate the isothermal
effectiveness factor for dynamic oxygen into mycelial pellets in
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial License, which permits use, distribution and reproduction in any
medium, provided the original work is properly cited and is not used for commercial purposes.
Kossen, 1977; Zhang & Zhang, 2016) were taken into account. In
brief, a coagulative pellet forms by aggregation of hundreds to
thousands of spores before they start to germinate, and a non-
coagulative pellet could form from the germination of a single
spore (Cairns et al., 2019). For the former case, we decided to
model the aggregation of spores by diffusion‐limited aggregation
(Witten & Sander, 1983; see Section 2). Figure 3 highlights that
the unbiased morphological simulations indeed enabled the de-
velopment of both coagulative and noncoagulative pellet types,
which reasonably matched the experimental data for the coagu-
lative A. niger MF19.5 and the noncoagulative R. stolonifer, re-
spectively. The growth process of a non‐coagulative pellet is
shown in Video S1.
F IGURE 2 Morphology and diffusivity of experimentally investigated pellets. Upper two rows: Exemplary projections of processedthree‐dimensional microcomputed tomography (μCT) images. Slices are from equatorial regions with a depth of 30 μm. Cubes are
50 × 50 × 50 μm. Bottom row: correlation between hyphal fraction (ch) and effective diffusion factor (keff ). Each blue data point corresponds toone cube applied for diffusion computations. The solid blue line is the correlation between the hyphal fraction and the effective diffusionfactor, resulting in the fitted exponent a in (Schmideder, Barthel, Müller, et al., 2019): k c1eff h
a= ( − ) . ± specifies the 95% confidence interval.μCT measurements were conducted with a voxel size of 1 μm (three Penicillium chrysogenum, three Aspergillus niger MF19.5, five
A. niger MF22.4, and three Rhizopus oryzae pellets) and 2 μm (11 R. stolonifer pellets). For each strain, at least 149 cubes were investigated.Scale bar: 100 μm [Color figure can be viewed at wileyonlinelibrary.com]
SCHMIDEDER ET AL. | 935
55
Pellet formation is not only dependent on coagulative and non-
coagulative spore aggregation type but also on the genetic make‐upof the species and process‐relevant parameters (Cairns et al., 2019).
To simulate a broad range of morphologically different pellets, our
simulation therefore considered high variation of five morphological
parameters (Figure 4): maximum growth angle, branch angle, hyphal
diameter, branch interval, and number of initial spores. For each of
these parameters, five values were estimated, resulting in 5 31255 =
morphologically different structures in total. In a recent study
(Lehmann et al., 2019), the investigation of 31 filamentous fungal
species resulted in mean branching angles between 26° and 86°, mean
internodal lengths (distance between two branches) between 40 and
453 μm, and mean hyphal diameters between 2.7 and 6.5 μm. How-
ever, the diameter of fungal hyphae has been reported to range from 2
to 10 μm (Meyer et al., 2020; Zacchetti et al., 2018). We decided to
also consider in the simulations the hyphal diameter of filamentous
bacteria (about 0.5–1 μm), as filamentous bacteria are morphologically
similar to filamentous fungi (Nielsen, 1996; Zacchetti et al., 2018), with
Streptomycetes as important cell factories for antibiotic production
(Nepal & Wang, 2019). In our simulations, the parameter branch angle
was set to 20°, 55°, 90°, 125°, or 160°, whereby angles larger than 90°
orient the branch towards the opposite direction of the leading
hyphae. The morphological parameter branch interval was defined as
the ratio between the internodal length and the hyphal diameter,
scaled with the hyphal diameter because large internodal lengths are
linked with large hyphal diameters (Lehmann et al., 2019). As
expected, simulations with large branch intervals produced disperse
mycelia instead of pellets. Thus, we set the maximum branch interval
F IGURE 3 Experimental and simulated coagulative and noncoagulative pellet formation. First and third rows show experimental data ofcoagulative pellet forming Aspergillus niger MF19.5 (Fiedler et al., 2018) and noncoagulative pellet forming Rhizopus stolonifer, respectively.
Second and fourth rows illustrate simulations of coagulative and noncoagulative pellet formation, respectively. First column shows spore(aggregates), second column germinated spore (aggregates), third column two‐dimensional (2D) projections of three‐dimensional (3D) pellets,and fourth column slices of pellets with a depth of 30 μm. Images of experimental spore aggregates were captured with light microscopy,whereas experimental pellets were measured with microcomputed tomography (μCT) to determine 3D images. Spore aggregation of coagulative
pellet formation was simulated with the diffusion‐limited aggregation method (Witten & Sander, 1983) and 2048 spores. Pellet growth wassimulated with the Monte Carlo model described in Section 2. Scale bar: 100 μm
936 | SCHMIDEDER ET AL.
56
to 35. The maximum number of initial spores was set to 10,000 be-
cause coagulative pellets can result from hundreds or thousands of
agglomerated spores (Fontaine et al., 2010; Metz & Kossen, 1977).
Due to the lack of literature describing the growth angle of hyphae, we
varied the maximum growth angle within a range to consider straight
growing (0°) and strongly curved (48°) hyphae.
Our Monte Carlo simulations covered a broad morphological
range of filamentous fungal pellets, which we believe also includes
pellets reflecting macromorphologies from smaller filamentous
bacteria.
Because image resolution could influence computed properties
(Schmideder, Barthel, Müller, et al., 2019; Velichko et al., 2009), we
further investigated the influence of pellet resolution on their dif-
fusivity. In Figure 5a, an exemplary cube of a simulated pellet is
shown that differs only in resolution, that is, the number of voxels
that span a given hyphal diameter. In this figure, the image resolution
increases from top to bottom, and the hyphae appear more cylind-
rical. Notably, the a value in k c1eff ha= ( − ) decreases with increasing
resolution of three exemplary simulated pellets and converges to
a1.6 1.8< < , suggesting that the a value converges with increasing
resolution.
To investigate the influence of the image resolution of experi-
mentally determined pellets on their diffusivity, we thus conducted
μCT measurements with different resolutions. In particular, R. stolo-
nifer pellets were measured with a voxel size of 1, 2, 3, and 4 μm and
R. oryzae pellets with a voxel size of 1 and 2 μm. The decrease in the
voxel size led to an increased resolution. Similar to simulated pellets,
a in k c1eff ha= ( − ) decreased with increasing resolution (Figure 5b).
Note that (i) measurements with a voxel size of 1 μm in the case of
R. stolonifer, the organism with the highest hyphal diameter in this
study, result in hollow hyphae (Figure S4), which cannot be used for
diffusion computations and (ii) that low resolutions result in coarse
surfaces of the hyphae that do not reflect their smooth nature. Thus,
3D images of filamentous pellets are limited to a minimum and
maximum resolution for each organism, which strongly depends on
their hyphal diameters. We therefore propose the application of an
image resolution that represents hyphae with about four to six voxels
in diameter.
Tomadakis and Robertson (2005) summarized and extended
correlations between the morphology and the diffusivity, con-
ductivity, and permeability through fiber structures with different
orientations. The study was based on previous studies by Tomadakis
& Sotirchos, (1991, 1993, 1993, 1993). To validate our diffusion
computations and to identify an appropriate resolution for simulated
pellets, we compared the solution for randomly orientated over-
lapping straight fibers by Tomadakis and Sotirchos (1991) with our
computed diffusivity through such a morphology. We set up a si-
mulation with 1000 spores, a growth angle of 0°, and allowed colli-
sions (Figure 6a) to obtain randomly orientated overlapping straight
fibers. As shown in Figure 6b,c, our simulated structure was
F IGURE 4 Applied morphological parameters for simulations of filamentous structures. Five morphological parameters were applied:
maximum growth angle, branch angle, hyphal diameter, branch interval (ratio between the distance between two branches and the hyphaldiameter), and number of initial spores. Each of these parameters was varied to five values resulting in a total of 5 31255 = simulationsperformed with the Monte Carlo method described in Section 2
SCHMIDEDER ET AL. | 937
57
comparable to the structure used for the correlation
keff0.037
1 0.037
0.661ε
ε( )=−
−(Tomadakis & Sotirchos, 1991). With increasing
resolution, that is, with increasing voxels representing the hyphal
diameter, the computed diffusivities of the simulated structure ap-
proached the correlation of Tomadakis and Sotirchos (1991;
Figure 6d). Especially in the hyphal fraction range 0–0.4, our com-
puted diffusivities fit well with their correlation, suggesting that the
diffusion computations were accurate for high resolutions in the
hyphal fraction range 0–0.4. The range reflects the hyphal fraction of
both simulated and experimentally determined pellets (Figures 2
and 7, see next Section 3.3).
3.3 | Merging diffusive mass transport data fromsimulated and µCT measured pellets
We investigated the diffusive mass transport in filamentous fungal
mycelia through the correlation between the diffusivity and mor-
phology of 3125 simulated and 66 μCT analyzed pellets. We applied
an image resolution that represented simulated hyphae five voxels in
diameter to make the results of simulations and experiments com-
parable (Figure 7a). Because this resolution could underestimate the
diffusivity (Figure 6 and our previous study; Schmideder, Barthel,
Müller, et al., 2019), we also applied a resolution that represented
hyphae with 13 voxels in diameter (Figure 7b). We considered this
resolution as a compromise between accuracy and computational
effort of diffusion computations. Similar to experimentally de-
termined pellets, we used several cubes of each simulated structure
for diffusion computations.
Interestingly, pellet formation did not occur for all parameter
combinations. The combination of low spore numbers and rare
branches resulted in dispersed mycelia instead of pellets. Thus, we
marked loose structures with a mean hyphal fraction of the re-
spective cubes less than 0.05. As a result, the initial 3125 simulated
structures were reduced to 1280 and 1791 pelletised structures for
hyphae represented with 5 and 13 voxels, respectively (Figure 7). The
mean correlation factor a aσ± ( ) of all 3125 structures was
a 2.059 0.182= ± and a 1.757 0.150= ± for hyphae represented
with 5 and 13 voxels, respectively (Figure 7, top and bottom). For
pelletized structures, the mean was altered to a 2.063 0.044= ±
and a 1.753 0.035= ± , respectively. In Figure 7, pellets with unlikely
morphological parameters were marked, namely all pellets with
straight hyphae (maximum growth angle 0= °) or extreme branch
angles (branch angle 20= ° and branch angle 160= °). Thus, the
F IGURE 5 Relationship between image resolution and effective diffusivity. Image resolution is represented by the number of voxels thatspan a given hyphal diameter. Left: Identical exemplary cube of a simulated pellet that differs only in image resolution. Exemplarily, hyphae
are represented with 3, 7, and 11 voxels in diameter. Right: a value of experimentally determined pellets and three exemplary simulated pelletsfitted on a base of at least 37 cubes per pellet with k c1eff h
a= ( − ) , where keff is the effective diffusion factor, ch the hyphal fraction,and a the fitted correlation factor (Schmideder, Barthel, Müller, et al., 2019). Morphological simulations of pellets were performed with the
Monte Carlo method described in the Section 2. Rhizopus pellets were measured with microcomputed tomography (μCT) with a voxelsize of 1 μm (three R. oryzae pellets), 2 μm (21 R. oryzae and 11 R. stolonifer pellets), 3 μm (13 R. stolonifer pellets), and 4 μm(seven R. stolonifer pellets). Each data point of measured pellets was fitted on the base of cubes of the mentioned number of pellets.
Error bars specify the 95% confidence interval [Color figure can be viewed at wileyonlinelibrary.com]
938 | SCHMIDEDER ET AL.
58
number of pellets was reduced to 898 and 712, with a narrow
distribution of the correlation factor a 2.072 0.025= ± and
a 1.760 0.023= ± for hyphae represented with 5 and 13 voxels in
diameter, respectively. Pellets with unlikely morphological para-
meters explain the outliers for the correlation factor a. For the ex-
perimentally determined pellets of P. chrysogenum, A. niger MF19.5,
A. nigerMF22.4, R. stolonifer, and R. oryzae, a was 2.05, 2.07,1.99,1.99,
and 1.99 (Figure 2), respectively, and thus in the range of the simu-
lated pellets where the hyphae are represented with 5 voxels in
diameter (Figure 7, top). Applying the maximum ball method
(Section 2), this is a mean hyphal diameter of 4.5 voxels that lies in
the range of our experimental pellets (4.3, 4.0, 4.3, 4.2, and 4.5 voxels
for the different strains).
The results strongly indicate that the diffusion law k c1eff ha= ( − )
is applicable to fungal pellets with arbitrary morphology. In addition,
the only fitting parameter a lies in a narrow range while the resolution
of the images does not change. This implies that there is a
generalizable law for the diffusivity through fungal pellets with the
hyphal fraction ch as the only independent variable. Contrary to our
expectations, detailed morphological parameters such as growth
angles, branch angles, hyphal diameters, number of initial spores, and
branching frequency did not affect the diffusive hindrance. For all
3125 simulated pellets we applied image resolutions to represent the
hyphal diameter with 5 and 13 voxels, respectively. We were able to
show that low resolutions (Figure 7a) lead to an underestimation of
the diffusivity (Figure 6 and our previous study; Schmideder, Barthel,
Müller, et al., 2019). In addition, the results of three simulated pellets
with different image resolutions (Figure 5b) show that the a parameter
converges with increasing resolution for each pellet. Thus, we suggest
to use the law for high resolutions (the hyphal diameter is represented
with 13 voxels in diameter; Figure 7b) to calculate the diffusivity
through fungal pellets:
k c1eff h1.76= ( − )
(3)
F IGURE 6 Validation of diffusion computations. The solution for the effective diffusion factor keff through randomly orientated overlappingstraight fibers keff
0.037
1 0.037
0.661ε
ε( )=−
−by Tomadakis and Sotirchos (1991) was compared with the computed keff through similar simulated structures.
The porosity was 1 hyphalfractionε = − . We performed the morphological simulations with the Monte Carlo method described in the Methods
section. (a) Whole simulated structure. (b) Exemplary cube of simulated structure. For diffusion computations, 76 cubes were investigated for eachresolution. Resolution is represented by the number of voxels that spanned the given hyphal diameter. (c) Model structure applied for solution byTomadakis and Sotirchos (1991). (d) Solution by Tomadakis and Sotirchos (1991) (black line) and computed diffusivities through simulated structure.
Each data point corresponds to the diffusivity through one cube [Color figure can be viewed at wileyonlinelibrary.com]
SCHMIDEDER ET AL. | 939
59
where keff is the effective diffusion factor (keff1− is similar to the for-
mation factor) describing the geometrical diffusion hindrance
through a 3D structure, and ch is the hyphal fraction (similar to solid
fraction). As shown in Figure S5, our law is not very distinct from the
law of random orientated overlapping fibers by Tomadakis and So-
tirchos (1991). However, our correlation differs very strongly from
the correlation for random orientated overlapping fibers proposed by
He et al. (2017). For low porosities, the correlation of He et al. (2017)
proposes effective diffusion factors larger than 1, which is physically
unreasonable. It has to be mentioned that our law might slightly
underestimate the diffusivity through filamentous fungal networks,
because Figure 5 implies that the convergence might not be fully
developed for a resolution of 13 voxels. However, doubling the re-
solution increases the volume of pellets and cubes for diffusion
computations eightfold. The increase of the cube‐volumes would
result in excessive computational times (With a resolution of 13
voxels, the diffusion computations of all 3125 pellets last about 3
weeks with an Intel Xeon E5‐1660 CPU (3.7 GHz)). In addition, some
pellets could not be reconstructed with very high resolutions because
they would result in working memories larger than our available
128 GB. Thus, the resolution of 13 voxels was a compromise between
accuracy and computational effort/feasibility.
4 | CONCLUSIONS
This study showed that a universal law (see Equation 3) holds for the
diffusive transport of nutrients, oxygen, and secreted metabolites
through any filamentous fungal pellet. As mentioned above, the
correlation is also known as Archie's law (Archie, 1942) with a ce-
mentation parameter equal to 1.76, which is in the range of porous
rocks (Glover et al., 1997, e.g., determined that the cementation
parameter of sandstone is between 1.5 and 2.5). Strictly speaking,
our law is valid for hyphal fractions less than 0.4 because the max-
imum hyphal fraction of simulated and measured pellets was 0.4.
However, to the best of our knowledge, there are no studies about
filamentous fungi where the measured hyphal fraction is more than
0.4. For example, Cui et al. (1997, 1998) reported average hyphal
F IGURE 7 Correlation factor a for simulated
and experimentally determined pellets. Each datapoint corresponds to the correlation between thehyphal fraction ch and the effective diffusion
factor keff of one pellet, resulting in the fittedexponent a. For each pellet, the fit is based ondiffusion computations through several cubes and
the correlation k c1effa
h= ( − ) . Hyphae ofsimulated pellets are represented with a“resolution” of five (top) and 13 (bottom) voxels indiameter. The increasing resolution decreases the
correlation factor a, and thus the diffusivitydecreases as well. Morphological simulations fordiffusion computations of simulated pellets were
performed with the Monte Carlo methoddescribed in the Section 2. Microcomputedtomography (μCT) measurements for diffusion
computations of experimentally determinedpellets were conducted with a voxel size of 1 μm(three Penicillium chrysogenum, three Aspergillus
niger MF19.5, five A. niger MF22.4, three Rhizopusoryzae pellets), and 2 μm (11 R. stolonifer pellets)[Color figure can be viewed atwileyonlinelibrary.com]
940 | SCHMIDEDER ET AL.
60
fractions between 0.07 and 0.30 for whole A. awamori pellets. If fu-
ture studies are confronted with hyphal fractions more than 0.4, we
propose to fall back on the correlation for randomly orientated
overlapping fibers proposed by Tomadakis and Sotirchos (1991),
which is (as already stated) not very distinct from our correlation.
They were able to show that their correlation was valid for solid
fractions less than 0.6, whereas higher solid fractions resulted in
another correlation. Our law predicts that only one independent
variable, the hyphal fraction (ch), affects pellet diffusivity. Knowing
the profile of the hyphal fraction in pellets, the law k c1eff h1.76= ( − )
determines the mass transport of molecules inside pellets. Molecule
concentrations inside pellets are affected by metabolic rates and the
transport through filamentous mycelia (Celler et al., 2012; Cui et al.,
1998). As the transport can now be calculated, the estimation of
metabolic rates within pellets becomes, for the first time, feasible.
Future experiments should provide information about the morphol-
ogy and metabolic activity or nutrient profiles of pellets, which can
be, for example, measured through flow cytometry or confocal laser
scanning microscopy (Schrinner et al., 2020; Tegelaar et al., 2020;
Veiter & Herwig, 2019). One possibility to determine the distribution
of oxygen and hyphal material inside pellets would be the application
of microelectrodes inside pellets (Hille et al., 2005; Wittier et al.,
1986) followed by μCT measurements (Schmideder, Barthel,
Friedrich, et al., 2019).
This generalized law was deduced from experimental and simu-
lation data. On the one hand, it proved the usefulness and power of
laboratory μCT systems to investigate the natural 3D morphology of
different filamentous fungi in utmost detail. However, as this method
is time and cost intensive and only applicable to a small number of
pellets, the use of Monte Carlo simulations proved to be powerful for
the massive generation of data covering a broad range of morpho-
logical characteristics. This computational approach allowed us to
generate a database of 3125 morphologically different pellets, which
will also open up new avenues of research. Such a database, which
can be accessed by the community, could also open new paths to-
wards parameter estimation of measured pellets of fungal or bac-
terial origin. Conclusions about the evolution of measured pellets
could be deduced because growth and morphological parameters of
simulated pellets are known.
We furthermore anticipate that this contribution will inspire
more sophisticated correlation measurements between morphology
and mass transport in any complex material, for example, in biofilms,
fiber materials, porous media, and fuel cells. In fact, correlations
between the morphology and transport properties (Archie, 1942;
Carman, 1937; Epstein, 1989; Kozeny, 1927) are crucial to compute
transport phenomena in several fields. Exemplarily, correlation laws
for fibrous materials (He et al., 2017; Tamayol et al., 2012; Tomadakis
& Robertson, 2005) are already applied to design nano‐ and micro-
porous membranes (Yuan et al., 2008), heat insulations (Panerai et al.,
2017) and dissipations (Jung et al., 2016), acoustic insulations (Tang
& Yan, 2017), electrodes (Kim et al., 2019), and batteries (Ke et al.,
2018). Concerning fibrous materials, a direct link to our study was
shown (Figure 6). For nonfibrous materials, a combination of 3D
image acquisition, morphological simulations, and mass transport
computations could significantly enhance the understanding of
morphology‐dependent transport phenomena. Hence, we foresee
that the biological and computational approach followed in this study
may be synergistically adopted to other research questions in bio
(techno)logy and beyond.
ACKNOWLEDGMENTS
Special thanks go to Katherina Celler and Gilles P. van Wezel and his
chair who provided the code used in their study (Celler et al., 2012). The
authors thank The Anh Baran for preliminary studies on the morpho-
logical simulations, Clarissa Schulze and Michaela Thalhammer for as-
sistance with μCT measurements, Andrea Pape for preparation of
pellets, Vincent Bürger for the preliminary code for diffusion‐limited
aggregation, and Markus Betz and Christian Preischl for preliminary
studies on diffusion computations. We also wish to thank Christoph
Kirse, Michael Kuhn, Johann Landauer, Thomas Riller, and Johannes
Petermeier for helpful and fruitful discussions. This study made use of
equipment that was funded by the Deutsche Forschungsgemeinschaft
(DFG, German Research Foundation)–198187031. The authors thank
the Deutsche Forschungsgemeinschaft for financial support for this
study within the SPP 1934 DiSPBiotech–315384307 and 315305620
and SPP2170 InterZell–427889137. Open access funding enabled and
organized by Projekt DEAL.
CONFLICTS OF INTEREST
The authors have no conflicts of interest to declare.
AUTHOR CONTRIBUTIONS
Stefan Schmideder did the conception and design of the study. Stefan
Schmideder, Henri Müller, and Lars Barthel wrote the manuscript,
which was edited and approved by all authors. Heiko Briesen and
Vera Meyer supervised the study. Lars Barthel and Ludwig Niessen
cultivated filamentous fungi and prepared pellets for microcomputed
tomography (μCT) measurements. Henri Müller and Stefan Schmi-
deder performed μCT measurements of pellets. Stefan Schmideder,
Henri Müller, and Tiaan Friedrich performed image analysis. Stefan
Schmideder and Tiaan Friedrich set up the code for morphological
Monte Carlo simulations. Stefan Schmideder performed the diffusion
computations and analyzed the results.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the
corresponding author upon reasonable request.
ORCID
Stefan Schmideder http://orcid.org/0000-0003-4328-9724
Henri Müller https://orcid.org/0000-0002-4831-0003
Lars Barthel https://orcid.org/0000-0001-8951-5614
Tiaan Friedrich https://orcid.org/0000-0001-8346-4908
Ludwig Niessen https://orcid.org/0000-0003-4083-2779
Schmideder, S., Barthel, L., Friedrich, T., Thalhammer, M., Kovačević, T.,Niessen, L., Meyer, V., & Briesen, H. (2019). An X‐raymicrotomography‐based method for detailed analysis of the three‐dimensional morphology of fungal pellets. Biotechnology and
Additional supporting information may be found online in the
Supporting Information section.
How to cite this article: Schmideder S, Müller H, Barthel L,
et al. Universal law for diffusive mass transport through
mycelial networks. Biotechnology and Bioengineering. 2021;
118:930–943. https://doi.org/10.1002/bit.27622
SCHMIDEDER ET AL. | 943
63
Supplementary Materials for Paper III: Universal law for
diffusive mass transport through mycelial networks
Fig. S1: Influence of cube size on diffusion computations
Diffusion computations were conducted with different cube sizes of experimentally determined pellets
and three exemplary simulated pellets. 𝑎 value was fitted on base of 𝑘𝑒𝑓𝑓 = (1 − 𝑐ℎ)𝑎, where 𝑘𝑒𝑓𝑓 is
the effective diffusion factor, 𝑐ℎ the hyphal fraction, and a the fitted correlation factor (Schmideder et
al., 2019). Error bars specify the 95% confidence interval. Morphological simulations of pellets were
performed with the Monte Carlo method described in the Methods section. Microcomputed
tomography measurements were conducted with a voxel size of 1 µm (three Penicillium
chrysogenum, three Aspergillus niger MF19.5, five Aspergillus niger MF22.4, three Rhizopus oryzae
pellets) and 2 µm (11 Rhizopus stolonifer pellets). Red circles mark the cube size used for diffusion
computations in the Results section of this study.
64
Fig. S2: Influence of cube size on porosity to determine size of representative elementary
volume (REV)
We investigated the porosity at four positions per pellet. At these positions we altered the cube size and
analyzed the porosity. Analyses were conducted with one pellet per experimentally determined strain
and three exemplary simulated pellets. Morphological simulations of pellets were performed with the
Monte Carlo method described in the Methods section. Microcomputed tomography measurements
were conducted with a voxel size of 1 μm (three Penicillium chrysogenum, three Aspergillus niger
MF19.5, five Aspergillus niger MF22.4, three Rhizopus oryzae pellets) and 2 μm (11 Rhizopus
stolonifer pellets).
65
Fig. S3: Influence of cube size on effective diffusion factor to determine size of representative
elementary volume (REV)
We investigated the effective diffusion factor at four positions per pellet. At these positions we altered
the cube size and analyzed the effective diffusion factor. Analyses were conducted with one pellet per
experimentally determined strain and three exemplary simulated pellets. Morphological simulations of
pellets were performed with the Monte Carlo method described in the Methods section. Microcomputed
tomography measurements were conducted with a voxel size of 1 μm (three Penicillium chrysogenum,
three Aspergillus niger MF19.5, five Aspergillus niger MF22.4, three Rhizopus oryzae pellets) and 2
μm (11 Rhizopus stolonifer pellets).
66
Fig. S4: Comparison of microcomputed tomography (µCT) images of R. stolonifer with a voxel
size of 1 and 2 µm.
Raw and binarised images are exemplary 250 × 250 µm plane cut outs of pellets. The slices of binarised
images are 250 × 250 × 10 µm of the same region as the raw and binarised images. Arrows indicate
exemplary hollow hyphae. Scale bar: 50 µm.
67
Fig. S5: Comparison of correlations between diffusivity and solid fraction.
The correlation proposed by He et al. (2017) and Tomadakis and Sotirchos (1992) were both proposed
for random orientated overlapping fibers. The effective diffusion factors 𝑘𝑒𝑓𝑓 are shown as functions
of the porosity 𝜀, which is defined by the solid/hyphal fraction 𝛷: 𝜀 = 1 − 𝛷.
68
1
Table S1: Parameters for morphological simulations. Parameters with the value ‘user defined’ could
be varied by the user and define all further parameters.
Parameter Symbol Value Unit
Number of spores 𝑁𝑠𝑝 User defined −
Hyphal diameter 𝑑ℎ User defined µ𝑚
Maximum growth angle 𝜃𝑔,𝑚𝑎𝑥 User defined °
Branch angle 𝜃𝑏 User defined °
Branch interval 𝑏 User defined −
Tip extension rate 𝛼𝑡 User defined µ𝑚 ℎ−1
Cultivation period 𝑡𝑒𝑛𝑑 User defined ℎ
Diameter of spores 𝑑𝑠𝑝 1.5 ∙ 𝑑ℎ µ𝑚
Length of germlings 𝑙𝑔𝑒𝑟𝑚 0.5 ∙ 𝑑𝑠𝑝 + 𝑑ℎ µ𝑚
Time step for calculation ∆𝑡 𝑑ℎ
2 𝛼𝑡
ℎ
Length of a segment 𝑙𝑠𝑒𝑔𝑚𝑒𝑛𝑡 𝑑ℎ
2
µ𝑚
Length of apical region 𝑙𝑎𝑝𝑖𝑐𝑎𝑙 𝑏 𝑑ℎ
4
µ𝑚
Length of subapical region 𝑙𝑠𝑢𝑏𝑎𝑝𝑖𝑐𝑎𝑙 𝑏 𝑑ℎ
4
µ𝑚
Minimum mean distance
between two branches
𝑑𝑏,𝑚𝑖𝑛 𝑏 𝑑ℎ µ𝑚
Hyphal diameter in voxels 𝑑ℎ,𝑣𝑥 User defined 𝑣𝑜𝑥
Scaling factor for nodes 𝑓𝑠𝑐𝑎𝑙𝑒 𝑑ℎ,𝑣𝑥
𝑑ℎ
𝑣𝑜𝑥 µ𝑚−1
69
Supplementary Protocol 1: Preparation of pellets
Aspergillus niger Rhizopus stolonifer
Initial cultivation Complete medium (CM) agar Potato extract glucose agar
Spore harvest Add 10 mL physiological salt (PS) solution, scrape spores off with sterile cotton stick and transfer to sterile 15 mL tube. Vortex for 30 s, filter through miracloth filter paper into fresh tube.
Add 10 mL physiological salt (PS) solution, scrape spores off with sterile cotton stick and transfer to sterile 15 mL tube. Vortex for 30 s, filter through miracloth filter paper into fresh tube.
Serial dilution 10-1 in sterile PS solution 10-1 in sterile PS solution
Spore count From 10-1 dilution, Thoma type counting chamber, 0.1 mm chamber depth.
From 10-1 dilution, Thoma type counting chamber, 0.1 mm chamber depth.
Storage of spores 4 °C 4 °C
Pellet production 50 mL CM medium in 250 mL shake flask
100 mL Potato extract glucose medium in 500 mL shake flask
Inoculation 5*106 spores per mL medium 8*102 spores per mL medium
Incubation conditions 250 rpm, 30 °C, 48 h 250 rpm, 23 °C, 48 h
PS solution Sodium chloride 8.9g Add aqua demin. to 1 L Autoclave at 121 °C for 20 min Preparation of CM medium ASP+N (50x), pH 5,5 20 mL Glucose solution (50% (w/v)) 20 mL Magnesium sulfate solution (1 M) 2 mL Trace element solution (1000x) 1 mL Casamino acid solution (10% (w/v)) 10 mL Yeast extract solution (10% (w/v) 50 mL Add aqua demin. to 1 L Autoclave at 121 °C for 20 min Preparation of ASP+N (50x) Potassium chloride 26.1 g Potassium dihydrogen phosphate 74.85 g Sodium nitrate 297.47 g Add aqua demin. to 1 L, pH 5.5 Autoclave at 121 °C for 20 min
70
Preparation of trace element solution (1000x) Zinc sulfate heptahydrate 12.27 g Boric acid 11 g Manganese (II) chloride tetrahydrate 3.15 g Iron (II) sulfate heptahydrate 2.73 g Cobalt (II) chloride hexahydrate 0.92 g Copper (II) sulfate pentahydrate 1.02 g Sodium molybdate dihydrate 1.28 g EDTA 50.85 g Add aqua demin. to 1 L Autoclave at 121 °C for 20 min Preparation of potato extract glucose Potato Extract Glucose Broth (Carl Roth) 26.5g Add aqua demin. to 1 L Autoclave at 121 °C for 20 min
71
Penicillium chrysogenum Rhizopus oryzae
Initial cultivation 3 % malt extract agar (MEA) 3 % MEA
Spore harvest 5 mL sterile tap water, 2 min. homogenization with sterile spatula. Fill to 15 mL sterile Falcon tube and centrifuge 5 min at 5,000 x g. Wash 2x with 5 mL sterile tap water and discard supernatant. Suspend washed conidiospores in 1 mL sterile tap water.
5 mL sterile tap water, 2 min. homogenization with sterile spatula. Fill to 15 mL sterile Falcon tube and centrifuge 5 min at 5,000 x g. Wash 2x with 5 mL sterile tap water and discard supernatant. Suspend washed sporangiospores in 1 mL sterile tap water.
Serial dilution 10-1-10-6 in sterile tap water 10-1-10-6 in sterile tap water
Spore count From 10-2 dilution, Thoma type counting chamber, 0.1 mm chamber depth.
From 10-2 dilution, Thoma type counting chamber, 0.1 mm chamber depth.
Storage of spores 4 °C 4 °C
Pellet production 100 mL pellet production medium for P. chrysogenum
100 mL pellet production medium for Rh. oryzae
Inoculation 104 spores per mL medium 105 spores per mL medium
Incubation conditions 250 rpm, 23 °C, 48 h 250 rpm, 23 °C, 16 h
Preparation of MEA Malt extract 30 g Peptone from soy 3 g Agar 15 g Add aqua demin. to 1 L, pH 5.2 Autoclave at 121 °C for 15 min Preparation of pellet production medium for P. chrysogenum Component 1: Ammoniumsulfat 0.55 % (w/v) autoclave at 121 °C for 15 min Component 2: Difco Yeast Carbon Base 11.7 % (w/v) Glucose 5,5 % (w/v) Filter sterilize though 0.2 µm membrane Mix Component 1: Component 2 9:1 prior to inoculation Preparation of pellet production medium for R. oryzae Glucose 5g Peptone from soy 1.5g Yeast extract 1.4g NaCl 2g K2HPO4 1g Add aqua demin. to 1 L, pH 5.0 Autoclave at 121 °C for 15 min
72
Supplementary Protocol 2:
X-ray microcomputed tomography (µCT) measurements of
filamentous fungal pellets
Equipment
Freeze-dried samples of filamentous fungal pellets (P. chrysogenum, A. niger, R. stolonifer,
and R. oryzae)
X-ray microcomputed tomography system (XCT‐1600HR; Matrix Technologies, Feldkirchen,
Germany)
Sample rod for µCT measurements with a voxel size of 1 µm
Sample rod for µCT measurements with a voxel size of 2 µm, 3 µm, and 4 µm
Self-constructed sample holder A made of polyoxymethylene (POM), in the shape of a
platform
Self-constructed sample holder B made of polyoxymethylene (POM), in the shape of a ton
Self-constructed sample holder C made of polyoxymethylene (POM) and polyimide (PI), in
the shape of a platform (POM) with an attached tube (PI)
King, 1998) should be applied to compute substrate-limited and substrate-saturated regions
of each pellet of the database. For the reaction term, substrate consumption rates have to be
acquired from literature or experiments. The spatial distribution of the hyphal fraction (from
database) and the universal diffusion law described in Paper III determine the diffusion term.
Depending on whether substrate limitation or saturation is desired, a suitable pellet structure
can be selected from the database. Based on process and/or genetic engineering approaches,
the chosen pellet structure should then be realized in bioreactors. Although there have been
many advances in morphological engineering, the realization of suitable pellets is most likely
going to be an iterative process. For this, a detailed verification of the pellet morphology
can be conducted based on µCT measurements and 3D image analysis. Obviously, some of
the proposed steps of the idealized workflow have to be investigated and elaborated in much
more detail. However, this workflow shows that the developed methods and findings embed-
ded in this thesis are important steps towards the construction of filamentous fungal pellets as
optimized production hosts.
So far, this dissertation mainly addressed individual pellets. However, submerged cultures
of filamentous microorganisms can result in broad pellet size distributions (Kurt et al., 2018;
Schrinner et al., 2020) influencing the further growth and production behavior (King, 1998;
Wösten et al., 2013) as well as the rheology and mass transfer in the bioprocess (Bliatsiou
et al., 2020). Contrary to the microscopic and continuum models elaborated in Section 2.3.2,
population balance equations (PBEs) allow the modeling of the development of the mor-
phological heterogeneity (King, 1998). On the one hand, recent advances in measuring the
macromorphology of numerous pellets (Sections 2.1.2 and 2.1.3) might contribute to PBEs
that describe the development of the size distribution (Edelstein and Hadar, 1983; Tough
et al., 1995). On the other hand, micromorphological data lacks as input of future PBEs with
80
Figure 5: Idealized workflow for the construction of filamentous fungal pellets as optimizedproduction hosts.
81
a high structuredness. As proposed in Schmideder et al. (2018), population balance models
with a high structuredness can consider several morphological properties including the size
of pellets, the diameter of the shell layer supplied with oxygen, the total length of active and
inactive hyphae, and the total number of growing tips. The location of tips and hyphal mate-
rial can be determined based on µCT measurements and subsequent image analysis (Paper I).
While laboratory µCT systems enable the visualization of a few pellets with one measurement,
hundreds of pellets can be visualized based on synchrotron radiation within a few minutes
(Section 6.1). Further, active and inactive regions of pellets can be estimated based on exper-
imental approaches (Section 2.2.2) or diffusion-reaction equations (Celler et al., 2012; King,
1998). Thus, PBEs with a high morphological structuredness can be developed through the
use of micromorphological data (Paper I) and the proposed universal diffusion law (Paper III).
Such modeling approaches will deepen the mechanistic understanding of the development of
(pellet) heterogeneities including growth, breakage, and aggregation processes.
To conclude, the developed methods to determine the micromorphology (Paper I) and effec-
tive diffusivity (Paper II) of pellets and the universal diffusion law through hyphal networks
(Paper III) will open new paths towards morphological engineering to enhance productivities
in filamentous fungal biotechnology.
82
7 References
Antecka, A., Bizukojc, M., and Ledakowicz, S. (2016). Modern morphological engineering techniques for im-proving productivity of filamentous fungi in submerged cultures. World Journal of Microbiology and Biotech-nology, 32:193.
Archie, G. E. (1942). The electrical resistivity log as an aid in determining some reservoir characteristics. Trans-actions of the AIME, 146(01):54–62.
Barry, D. J., Williams, G. A., and Chan, C. (2015). Automated analysis of filamentous microbial morphologywith AnaMorf. Biotechnology Progress, 31(3):849–852.
Becker, J., Wieser, C., Fell, S., and Steiner, K. (2011). A multi-scale approach to material modeling of fuel celldiffusion media. International Journal of Heat and Mass Transfer, 54(7-8):1360–1368.
Biesebeke, R., Record, E., van Biezen, N., Heerikhuisen, M., Franken, A., Punt, P. J., and van den Hondel, C. A.M. J. J. (2005). Branching mutants of Aspergillus oryzae with improved amylase and protease production onsolid substrates. Applied Microbiology and Biotechnology, 69(1):44–50.
Bizukojc, M. and Ledakowicz, S. (2006). A kinetic model to predict biomass content for Aspergillus nigergerminating spores in the submerged culture. Process Biochemistry, 41(5):1063–1071.
Bizukojc, M. and Ledakowicz, S. (2010). The morphological and physiological evolution of Aspergillus terreusmycelium in the submerged culture and its relation to the formation of secondary metabolites. World Journalof Microbiology and Biotechnology, 26:41.
Bliatsiou, C., Schrinner, K., Waldherr, P., Tesche, S., Böhm, L., Kraume, M., and Krull, R. (2020). Rheologicalcharacteristics of filamentous cultivation broths and suitable model fluids. Biochemical Engineering Journal,163:107746.
Bocking, S. P., Wiebe, M. G., Robson, G. D., Hansen, K., Christiansen, L. H., and Trinci, A. P. J. (1999).Effect of branch frequency in Aspergillus oryzae on protein secretion and culture viscosity. Biotechnology andBioengineering, 65(6):638–648.
Böl, M., Schrinner, K., Tesche, S., and Krull, R. (2021). Challenges of influencing cellular morphology bymorphology engineering techniques and mechanical induced stress on filamentous pellet systems - A criticalreview. Engineering in Life Sciences, 21(3-4):51–67.
Brakhage, A. A. (2013). Regulation of fungal secondary metabolism. Nature Reviews Microbiology, 11(1):21–32.
Brunk, M., Sputh, S., Doose, S., van de Linde, S., and Terpitz, U. (2018). HyphaTracker: An ImageJ toolbox fortime-resolved analysis of spore germination in filamentous fungi. Scientific Reports, 8(1):605.
Buschulte, T. C. (1992). Mathematische Modellbildung und Simulation von Zellwachstum, Stofftransport undStoffwechsel in Pellets aus Streptomyceten. PhD thesis, Fakultät Verfahrenstechnik der Universität Stuttgart.
Cairns, T. C., Feurstein, C., Zheng, X., Zheng, P., Sun, J., and Meyer, V. (2019a). A quantitative image analysispipeline for the characterization of filamentous fungal morphologies as a tool to uncover targets for morphologyengineering: a case study using aplD in Aspergillus niger. Biotechnology for Biofuels, 12:149.
83
Cairns, T. C., Nai, C., and Meyer, V. (2018). How a fungus shapes biotechnology: 100 years of Aspergillus nigerresearch. Fungal Biology and Biotechnology, 5:13.
Cairns, T. C., Zheng, X., Zheng, P., Sun, J., and Meyer, V. (2019b). Moulding the mould: understanding andreprogramming filamentous fungal growth and morphogenesis for next generation cell factories. Biotechnologyfor Biofuels, 12(1):77.
Cardini, A., Pellegrino, E., Del Dottore, E., Gamper, H. A., Mazzolai, B., and Ercoli, L. (2020). HyLength:a semi-automated digital image analysis tool for measuring the length of roots and fungal hyphae of densemycelia. Mycorrhiza, 30(2-3):229–242.
Celler, K., Picioreanu, C., van Loosdrecht, Mark C. M., and van Wezel, G. P. (2012). Structured morphologi-cal modeling as a framework for rational strain design of Streptomyces species. Antonie van Leeuwenhoek,102(3):409–423.
Choy, V., Patel, N., and Thibault, J. (2011). Application of image analysis in the fungal fermentation of Tricho-derma reesei RUT-C30. Biotechnology Progress, 27(6):1544–1553.
Colin, V. L., Baigori, M. D., and Pera, L. M. (2013). Tailoring fungal morphology of Aspergillus niger MYA 135by altering the hyphal morphology and the conidia adhesion capacity: biotechnological applications. AMBExpress, 3:27.
Cronenberg, C. C. H., Ottengraf, S. P. P., Heuvel, J. C., Pottel, F., Sziele, D., Schgerl, K., and Bellgardt, K. H.(1994). Influence of age and structure of Pencillium chrysogenum pellets on the internal concentration profiles.Bioprocess Engineering, 10(5-6):209–216.
Cui, Y. Q., van der Lans, R., and Luyben, K. (1998). Effects of dissolved oxygen tension and mechanical forceson fungal morphology in submerged fermentation. Biotechnology and Bioengineering, 57(4):409–419.
Cui, Y. Q., vanderLans, R., and Luyben, K. (1997). Effect of agitation intensities on fungal morphology ofsubmerged fermentation. Biotechnology and Bioengineering, 55(5):715–726.
Demain, A. L. (2014). Importance of microbial natural products and the need to revitalize their discovery. Journalof Industrial Microbiology and Biotechnology, 41(2):185–201.
Driouch, H., Haensch, R., Wucherpfennig, T., Krull, R., and Wittmann, C. (2012). Improved enzyme productionby bio-pellets of Aspergillus niger: Targeted morphology engineering using titanate microparticles. Biotech-nology and Bioengineering, 109(2):462–471.
Driouch, H., Sommer, B., and Wittmann, C. (2010). Morphology Engineering of Aspergillus niger for ImprovedEnzyme Production. Biotechnology and Bioengineering, 105(6):1058–1068.
Du, H., Lv, P., Ayouz, M., Besserer, A., and Perré, P. (2016). Morphological characterization and quantifi-cation of the mycelial growth of the brown-rot fungus Postia placenta for modeling purposes. PloS ONE,11(9):e0162469.
Dynesen, J. and Nielsen, J. (2003). Surface hydrophobicity of Aspergillus nidulans conidiospores and its role inpellet formation. Biotechnology Progress, 19(3):1049–1052.
Edelstein, L. and Hadar, Y. (1983). A model for pellet size distributions in submerged mycelial cultures. Journalof Theoretical Biology, 105(3):427–452.
Ehgartner, D., Herwig, C., and Fricke, J. (2017). Morphological analysis of the filamentous fungus Penicilliumchrysogenum using flow cytometry-the fast alternative to microscopic image analysis. Applied Microbiologyand Biotechnology, 101(20):7675–7688.
84
El-Enshasy, H., Kleine, J., and Rinas, U. (2006). Agitation effects on morphology and protein productive frac-tions of filamentous and pelleted growth forms of recombinant Aspergillus niger. Process Biochemistry,41(10):2103–2112.
Emerson, S. (1950). The growth phase in Neurospora corresponding to the logarithmic phase in unicellularorganisms. Journal of Bacteriology, 60(3):221–223.
Fiedler, M. R. M., Barthel, L., Kubisch, C., Nai, C., and Meyer, V. (2018). Construction of an improved As-pergillus niger platform for enhanced glucoamylase secretion. Microbial Cell Factories, 17:95.
Fleming, A. (1929). On the antibacterial action of cultures of a penicillium, with special reference to their use inthe isolation of B. influenzae. British Journal of Experimental Pathology, 10(3):226–236.
Fontaine, T., Beauvais, A., Loussert, C., Thevenard, B., Fulgsang, C. C., Ohno, N., Clavaud, C., Prevost, M.-C.,and Latgé, J.-P. (2010). Cell wall alpha1-3glucans induce the aggregation of germinating conidia of Aspergillusfumigatus. Fungal Genetics and Biology, 47(8):707–712.
Gibbs, P. A., Seviour, R. J., and Schmid, F. (2000). Growth of filamentous fungi in submerged culture: Problemsand possible solutions. Critical Reviews in Biotechnology, 20(1):17–48.
Grimm, L. H., Kelly, S., Hengstler, J., Gobel, A., Krull, R., and Hempel, D. C. (2004). Kinetic studies on theaggregation of Aspergillus niger conidia. Biotechnology and Bioengineering, 87(2):213–218.
Grimm, L. H., Kelly, S., Krull, R., and Hempel, D. C. (2005). Morphology and productivity of filamentous fungi.Applied Microbiology and Biotechnology, 69(4):375–384.
He, R., Li, C., Ma, L., Zhang, D., and Chen, S. (2016). Effect of highly branched hyphal morphology on theenhanced production of cellulase in Trichoderma reesei DES-15. 3 Biotech, 6:214.
Hille, A., Neu, T. R., Hempel, D. C., and Horn, H. (2005). Oxygen profiles and biomass distribution in biopelletsof Aspergillus niger. Biotechnology and Bioengineering, 92(5):614–623.
Hille, A., Neu, T. R., Hempel, D. C., and Horn, H. (2009). Effective diffusivities and mass fluxes in fungalbiopellets. Biotechnology and Bioengineering, 103(6):1202–1213.
Hoffmeister, D. and Keller, N. P. (2007). Natural products of filamentous fungi: enzymes, genes, and theirregulation. Natural Product Reports, 24(2):393–416.
Karahalil, E., Coban, H. B., and Turhan, I. (2019). A current approach to the control of filamentous fungal growthin media: microparticle enhanced cultivation technique. Critical Reviews in Biotechnology, 39(2):192–201.
Ke, X., Prahl, J. M., Alexander, J. I. D., Wainright, J. S., Zawodzinski, T. A., and Savinell, R. F. (2018).Rechargeable redox flow batteries: flow fields, stacks and design considerations. Chemical Society Reviews,47(23):8721–8743.
Keller, N. P. (2019). Fungal secondary metabolism: regulation, function and drug discovery. Nature ReviewsMicrobiology, 17(3):167–180.
Kelly, S., Grimm, L. H., Jonas, R., Hempel, D. C., and Krull, R. (2006). Investigations of the morphogenesis offilamentous microorganisms. Engineering in Life Sciences, 6(5):475–480.
Kim, M. J., Seo, Y., Cruz, M. A., and Wiley, B. J. (2019). Metal nanowire felt as a flow-through electrode forhigh-productivity electrochemistry. ACS Nano, 13(6):6998–7009.
King, R. (1998). Mathematical modelling of the morphology of streptomyces species. In Relation betweenmorphology and process performances, pages 95–124. Springer, Berlin, Heidelberg.
85
King, R. (2015). A framework for an organelle-based mathematical modeling of hyphae. Fungal Biology andBiotechnology, 2:5.
Krull, R., Cordes, C., Horn, H., Kampen, I., Kwade, A., Neu, T. R., and Nörtemann, B. (2010). Morphologyof filamentous fungi: linking cellular biology to process engineering using Aspergillus niger. In BiosystemsEngineering II: Linking Cellular Networks and Bioprocesses, pages 1–21. Springer, Berlin, Heidelberg.
Krull, R., Wucherpfennig, T., Esfandabadi, M. E., Walisko, R., Melzer, G., Hempel, D. C., Kampen, I., Kwade,A., and Wittmann, C. (2013). Characterization and control of fungal morphology for improved productionperformance in biotechnology. Journal of Biotechnology, 163(2):112–123.
Kunz, P. J., Barthel, L., Meyer, V., and King, R. (2020). Vesicle transport and growth dynamics in Aspergillusniger: Microscale modeling of secretory vesicle flow and centerline extraction from confocal fluorescent data.Biotechnology and Bioengineering, 117(9):2875–2886.
Kurt, T., Marbà-Ardébol, A.-M., Turan, Z., Neubauer, P., Junne, S., and Meyer, V. (2018). Rocking Aspergillus:morphology-controlled cultivation of Aspergillus niger in a wave-mixed bioreactor for the production of sec-ondary metabolites. Microbial Cell Factories, 17:128.
Kwon, M. J., Nitsche, B. M., Arentshorst, M., Jørgensen, T. R., Ram, A. F. J., and Meyer, V. (2013). Thetranscriptomic signature of RacA activation and inactivation provides new insights into the morphogeneticnetwork of Aspergillus niger. PloS ONE, 8(7):e68946.
Lecault, V., Patel, N., and Thibault, J. (2007). Morphological characterization and viability assessment of Tricho-derma reesei by image analysis. Biotechnology Progress, 23(3):734–740.
Lehmann, A., Zheng, W., Soutschek, K., Roy, J., Yurkov, A. M., and Rillig, M. C. (2019). Tradeoffs in hyphaltraits determine mycelium architecture in saprobic fungi. Scientific Reports, 9:14152.
Lejeune, R. and Baron, G. V. (1997). Simulation of growth of a filamentous fungus in 3 dimensions. Biotechnologyand Bioengineering, 53(2):139–150.
Lin, P.-J., Grimm, L. H., Wulkow, M., Hempel, D. C., and Krull, R. (2008). Population balance modeling of theconidial aggregation of Aspergillus niger. Biotechnology and Bioengineering, 99(2):341–350.
Lin, P.-J., Scholz, A., and Krull, R. (2010). Effect of volumetric power input by aeration and agitation on pelletmorphology and product formation of Aspergillus niger. Biochemical Engineering Journal, 49(2):213–220.
Metz, B. and Kossen, N. W. F. (1977). The growth of molds in the form of pellets–a literature review. Biotech-nology and Bioengineering, 19(6):781–799.
Meyer, V. (2008). Genetic engineering of filamentous fungi - Progress, obstacles and future trends. BiotechnologyAdvances, 26(2):177–185.
Meyer, V., Andersen, M. R., Brakhage, A. A., Braus, G. H., Caddick, M. X., Cairns, T. C., de Vries, R. P.,Haarmann, T., Hansen, K., Hertz-Fowler, C., Krappmann, S., Mortensen, U. H., Peñalva, M. A., Ram, A. F. J.,and Head, R. M. (2016). Current challenges of research on filamentous fungi in relation to human welfare anda sustainable bio-economy: a white paper. Fungal Biology and Biotechnology, 3:6.
Meyer, V., Basenko, E., Benz, P., Braus, G., Caddick, M., Csukai, M., Vries, R., Endy, D., Frisvad, J., Gunde-Cimerman, N., Haarmann, T., Hadar, Y., Hansen, K., Johnson, R., Keller, N., Kraševec, N., Mortensen, U.,Perez, R., Ram, A., Record, E., Ross, P., Shapaval, V., Steiniger, C., van den Brink, H., van Munster, J., Yarden,O., and Wösten, H. (2020). Growing a circular economy with fungal biotechnology: a white paper. FungalBiology and Biotechnology, 7:5.
Meyerhoff, J., Tiller, V., and Bellgardt, K.-H. (1995). Two mathematical models for the development of a singlemicrobial pellet. Bioprocess and Biosystems Engineering, 12(6):305–313.
86
Müller, H., Schmideder, S., Barthel, L., Niessen, L., Meyer, V., and Briesen, H. (2020). Optimized X–raymicrocomputed tomography and 3D volumetric image processing of filamentous fungal pellets. ProcessNetJahrestagung und DECHEMA Jahrestagung der Biotechnologen 2020, Web-Konferenz.
Nielsen, J., Johansen, C. L., Jacobsen, M., Krabben, P., and Villadsen, J. (1995). Pellet formation and fragmenta-tion in submerged cultures of Penicillium chrysogenum and its relation to penicillin production. BiotechnologyProgress, 11(1):93–98.
Nielsen, J. C., Grijseels, S., Prigent, S., Ji, B., Dainat, J., Nielsen, K. F., Frisvad, J. C., Workman, M., andNielsen, J. (2017). Global analysis of biosynthetic gene clusters reveals vast potential of secondary metaboliteproduction in Penicillium species. Nature Microbiology, 2:17044.
Nieminen, L., Webb, S., Smith, M. C. M., and Hoskisson, P. A. (2013). A flexible mathematical model plat-form for studying branching networks: Experimentally validated using the model actinomycete, Streptomycescoelicolor. PloS ONE, 8(2):e54316.
Olmos, E., Mehmood, N., Haj Husein, L., Goergen, J.-L., Fick, M., and Delaunay, S. (2013). Effects of bioreactorhydrodynamics on the physiology of Streptomyces. Bioprocess and Biosystems Engineering, 36(3):259–272.
Panerai, F., Ferguson, J. C., Lachaud, J., Martin, A., Gasch, M. J., and Mansour, N. N. (2017). Micro-tomographybased analysis of thermal conductivity, diffusivity and oxidation behavior of rigid and flexible fibrous insula-tors. International Journal of Heat and Mass Transfer, 108:801–811.
Papagianni, M. (2004). Fungal morphology and metabolite production in submerged mycelial processes. Biotech-nology Advances, 22(3):189–259.
Papagianni, M. (2014). Characterization of fungal morphology using digital image analysis techniques. Journalof Microbial and Biochemical Technology, 6(4):189–194.
Papagianni, M. and Mattey, M. (2006). Morphological development of Aspergillus niger in submerged citric acidfermentation as a function of the spore inoculum level. Application of neural network and cluster analysis forcharacterization of mycelial morphology. Microbial Cell Factories, 5:3.
Papagianni, M., Mattey, M., and Kristiansen, B. (1999). The influence of glucose concentration on citric acidproduction and morphology of Aspergillus niger in batch and culture. Enzyme and Microbial Technology,25(8-9):710–717.
Paul, G. C., Kent, C. A., and Thomas, C. R. (1993). Viability testing and characterization of germination of fungalspores by automatic image analysis. Biotechnology and Bioengineering, 42(1):11–23.
Pearson, A. P., Glennon, B., and Kieran, P. M. (2003). Comparison of morphological characteristics of Strep-tomyces natalensis by image analysis and focused beam reflectance measurement. Biotechnology Progress,19(4):1342–1347.
Pearson, A. P., Glennon, B., and Kieran, P. M. (2004). Monitoring of cell growth using the focused beam re-flectance method. Journal of Chemical Technology and Biotechnology, 79(10):1142–1147.
Perrins, W. T., McKenzie, D. R., and Mc Phedran, R. C. (1979). Transport properties of regular arrays of cylinders.Proceedings of the Royal Society A, 369(1737):207–225.
Petersen, N., Stocks, S., and Gernaey, K. V. (2008). Multivariate models for prediction of rheological char-acteristics of filamentous fermentation broth from the size distribution. Biotechnology and Bioengineering,100(1):61–71.
Pirt, S. J. (1966). A theory of the mode of growth of fungi in the form of pellets in submerged culture. Proceedingsof the Royal Society B, 166(1004):369–373.
87
Posch, A. E., Herwig, C., and Spadiut, O. (2013). Science-based bioprocess design for filamentous fungi. Trendsin Biotechnology, 31(1):37–44.
Priegnitz, B.-E., Wargenau, A., Brandt, U., Rohde, M., Dietrich, S., Kwade, A., Krull, R., and Fleissner, A.(2012). The role of initial spore adhesion in pellet and biofilm formation in Aspergillus niger. Fungal Geneticsand Biology, 49(1):30–38.
Prosser, J. I. and Tough, A. J. (1991). Growth mechanisms and growth kinetics of filamentous microorganisms.Critical Reviews in Biotechnology, 10(4):253–274.
Punt, P. J., van Biezen, N., Conesa, A., Albers, A., Mangnus, J., and van den Hondel, C. (2002). Filamentousfungi as cell factories for heterologous protein production. Trends in Biotechnology, 20(5):200–206.
Quintanilla, D., Chelius, C., Iambamrung, S., Nelson, S., Thomas, D., Gernaey, K. V., and Marten, M. R. (2018).A fast and simple method to estimate relative, hyphal tensile-strength of filamentous fungi used to assess theeffect of autophagy. Biotechnology and Bioengineering, 115(3):597–605.
Riquelme, M., Aguirre, J., Bartnicki-García, S., Braus, G. H., Feldbrügge, M., Fleig, U., Hansberg, W., Herrera-Estrella, A., Kämper, J., Kück, U., Mouriño-Pérez, R. R., Takeshita, N., and Fischer, R. (2018). Fungalmorphogenesis, from the polarized growth of hyphae to complex reproduction and infection structures. Micro-biology and Molecular Biology Reviews, 82(2):e00068–17.
Rønnest, N. P., Stocks, S. M., Lantz, A. E., and Gernaey, K. V. (2012). Comparison of laser diffraction andimage analysis for measurement of Streptomyces coelicolor cell clumps and pellets. Biotechnology Letters,34(8):1465–1473.
Sachs, C. C., Koepff, J., Wiechert, W., Grünberger, A., and Nöh, K. (2019). mycelyso - high-throughput analysisof Streptomyces mycelium live cell imaging data. BMC Bioinformatics, 20:452.
Salvo, L., Cloetens, P., Maire, E., Zabler, S., Blandin, J., Buffière, J., Ludwig, W., Boller, E., Bellet, D., andJosserond, C. (2003). X-ray micro-tomography an attractive characterisation technique in materials science.Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms,200:273–286.
Schmideder, S., Barthel, L., Friedrich, T., Thalhammer, M., Kovacevic, T., Niessen, L., Meyer, V., and Briesen, H.(2019a). An X-ray microtomography-based method for detailed analysis of the three-dimensional morphologyof fungal pellets. Biotechnology and Bioengineering, 116(6):1355–1365.
Schmideder, S., Barthel, L., Müller, H., Meyer, V., and Briesen, H. (2019b). From three-dimensional morphologyto effective diffusivity in filamentous fungal pellets. Biotechnology and Bioengineering, 116(12):3360–3371.
Schmideder, S., Friedrich, T., Kovacevic, T., Barthel, L., Kunz, P., Meyer, V., King, R., and Briesen, H. (2018).Growth of filamentous microorganisms: PBM and experimental determination of rate equations. ConferenceProcceedings: 6th Population Balance Modelling Conference (PBM 2018), Ghent, Belgium.
Schmideder, S., Müller, H., Barthel, L., Friedrich, T., Niessen, L., Meyer, V., and Briesen, H. (2021). Universallaw for diffusive mass transport through mycelial networks. Biotechnology and Bioengineering, 118(2):930–943.
Schrinner, K., Veiter, L., Schmideder, S., Doppler, P., Schrader, M., Münch, N., Althof, K., Kwade, A., Briesen,H., Herwig, C., and Krull, R. (2020). Morphological and physiological characterization of filamentous Lentzeaaerocolonigenes: Comparison of biopellets by microscopy and flow cytometry. PloS ONE, 15(6):e0234125.
Silva, E. M. E., Gutierrez, G. F., Dendooven, L., Hugo, J. I., and Ochoa-Tapia, J. A. (2001). A method toevaluate the isothermal effectiveness factor for dynamic oxygen into mycelial pellets in submerged cultures.Biotechnology Progress, 17(1):95–103.
88
Stock, S. R. (2009). Microcomputed tomography: Methodology and applications. CRC Press.
Tang, X. and Yan, X. (2017). Acoustic energy absorption properties of fibrous materials: A review. CompositesPart A: Applied Science and Manufacturing, 101:360–380.
Tegelaar, M., Aerts, D., Teertstra, W. R., and Wösten, H. A. B. (2020). Spatial induction of genes encodingsecreted proteins in micro-colonies of Aspergillus niger. Scientific Reports, 10:1536.
Tomadakis, M. M. and Robertson, T. J. (2005). Viscous permeability of random fiber structures: comparison ofelectrical and diffusional estimates with experimental and analytical results. Journal of Composite Materials,39(2):163–188.
Tomadakis, M. M. and Sotirchos, S. V. (1991). Effects of Fiber Orientation and Overlapping on Knudsen, Tran-sition, and Ordinary Regime Diffusion in Fibrous Substrates. MRS Proceedings, 250:221–226.
Tough, A. J., Pulham, J., and Prosser, J. I. (1995). A mathematical model for the growth of mycelial pelletpopulations. Biotechnology and Bioengineering, 46(6):561–572.
Veiter, L. and Herwig, C. (2019). The filamentous fungus Penicillium chrysogenum analysed via flow cytometry-a fast and statistically sound insight into morphology and viability. Applied Microbiology and Biotechnology,103(16):6725–6735.
Veiter, L., Rajamanickam, V., and Herwig, C. (2018). The filamentous fungal pellet-relationship between mor-phology and productivity. Applied Microbiology and Biotechnology, 102(7):2997–3006.
Velichko, A., Wiegmann, A., and Mücklich, F. (2009). Estimation of the effective conductivities of complex castiron microstructures using FIB-tomographic analysis. Acta Materialia, 57(17):5023–5035.
Vidal-Diez de Ulzurrun, G., Huang, T.-Y., Chang, C.-W., Lin, H.-C., and Hsueh, Y.-P. (2019). Fungal featuretracker (FFT): A tool for quantitatively characterizing the morphology and growth of filamentous fungi. PLoSComputational Biology, 15(10):e1007428.
Vignoles, G. L., Coindreau, O., Ahmadi, A., and Bernard, D. (2007). Assessment of geometrical and transportproperties of a fibrous C/C composite preform as digitized by x-ray computerized microtomography: Part II.Heat and gas transport properties. Journal of Materials Research, 22(6):1537–1550.
Villena, G. K., Fujikawa, T., Tsuyumu, S., and Gutierrez-Correa, M. (2010). Structural analysis of biofilms andpellets of Aspergillus niger by confocal laser scanning microscopy and cryo scanning electron microscopy.Bioresource Technology, 101(6):1920–1926.
Walisko, J., Vernen, F., Pommerehne, K., Richter, G., Terfehr, J., Kaden, D., Dähne, L., Holtmann, D., and Krull,R. (2017). Particle-based production of antibiotic rebeccamycin with Lechevalieria aerocolonigenes. ProcessBiochemistry, 53:1–9.
Wang, B., Chen, J., Li, H., Sun, F., Li, Y., and Shi, G. (2017). Pellet-dispersion strategy to simplify the seedcultivation of Aspergillus niger and optimize citric acid production. Bioprocess and Biosystems Engineering,40(1):45–53.
Wang, Q., Zhong, C., and Xiao, H. (2020). Genetic engineering of filamentous fungi for efficient protein expres-sion and secretion. Frontiers in Bioengineering and Biotechnology, 8:293.
Ward, O. P. (2012). Production of recombinant proteins by filamentous fungi. Biotechnology Advances,30(5):1119–1139.
Whelan, J., Murphy, E., Pearson, A., Jeffers, P., Kieran, P., McDonnell, S., Raposo, S., Lima-Costa, M. E.,and Glennon, B. (2012). Use of focussed beam reflectance measurement (FBRM) for monitoring changes inbiomass concentration. Bioprocess and Biosystems Engineering, 35(6):963–975.
89
Willemse, J., Büke, F., van Dissel, D., Grevink, S., Claessen, D., and van Wezel, G. P. (2018). SParticle, analgorithm for the analysis of filamentous microorganisms in submerged cultures. Antonie van Leeuwenhoek,111(2):171–182.
Wittier, R., Baumgartl, H., Lübbers, D. W., and Schügerl, K. (1986). Investigations of oxygen transfer into Peni-cillium chrysogenum pellets by microprobe measurements. Biotechnology and Bioengineering, 28(7):1024–1036.
Wösten, H. A. B. (2019). Filamentous fungi for the production of enzymes, chemicals and materials. CurrentOpinion in Biotechnology, 59:65–70.
Wösten, H. A. B., van Veluw, G. J., de Bekker, C., and Krijgsheld, P. (2013). Heterogeneity in the mycelium:implications for the use of fungi as cell factories. Biotechnology Letters, 35(8):1155–1164.
Wucherpfennig, T., Hestler, T., and Krull, R. (2011). Morphology engineering - osmolality and its effect onAspergillus niger morphology and productivity. Microbial Cell Factories, 10:58.
Xin, B., Xia, Y., Zhang, Y., Aslam, H., Liu, C., and Chen, S. (2012). A feasible method for growing fungal pelletsin a column reactor inoculated with mycelium fragments and their application for dye bioaccumulation fromaqueous solution. Bioresource Technology, 105:100–105.
Yang, H., King, R., Reichl, U., and Gilles, E. D. (1992a). Mathematical model for apical growth, septation, andbranching of mycelial microorganisms. Biotechnology and Bioengineering, 39(1):49–58.
Yang, H., Reichl, U., King, R., and Gilles, E. D. (1992b). Measurement and simulation of the morphologicaldevelopment of filamentous microorganisms. Biotechnology and Bioengineering, 39(1):44–48.
Yaws, C. L. (2014). Diffusion coefficient at infinite dilution in Water - inorganic compounds. In transportproperties of chemicals and hydrocarbons, pages 704–705. William Andrew.
Yuan, J., Liu, X., Akbulut, O., Hu, J., Suib, S. L., Kong, J., and Stellacci, F. (2008). Superwetting nanowiremembranes for selective absorption. Nature Nanotechnology, 3(6):332–336.
Zacchetti, B., Wösten, H. A. B., and Claessen, D. (2018). Multiscale heterogeneity in filamentous microbes.Biotechnology Advances, 36(8):2138–2149.
Zhang, J. and Zhang, J. (2016). The filamentous fungal pellet and forces driving its formation. Critical Reviewsin Biotechnology, 36(6):1066–1077.
Žnidaršic, P., Komel, R., and Pavko, A. (1998). Studies of a pelleted growth form of Rhizopus nigricans as abiocatalyst for progesterone 11alpha-hydroxylation. Journal of Biotechnology, 60(3):207–216.
Žnidaršic, P., Komel, R., and Pavko, A. (2000). Influence of some environmental factors on Rhizopus nigricanssubmerged growth in the form of pellets. World Journal of Microbiology and Biotechnology, 16(7):589–593.
90
8 Appendix: List of Publications
Peer-Reviewed Publications
Stefan Schmideder, Christoph Kirse, Julia Hofinger, Sascha Rollié, and Heiko Briesen,
2018, Modeling the separation of microorganisms in bioprocesses by flotation. Pro-
cesses, 6(10): 184.
Stefan Schmideder, Lars Barthel, Tiaan Friedrich, Michaela Thalhammer, Tijana Ko-
vacevic, Ludwig Niessen, Vera Meyer, and Heiko Briesen, 2019, An X-ray microtomography-
based method for detailed analysis of the three-dimensional morphology of fungal pel-
lets. Biotechnology and Bioengineering, 116(6): 1355–1365.
Stefan Schmideder, Lars Barthel, Henri Müller, Vera Meyer, and Heiko Briesen, 2019,
From three-dimensional morphology to effective diffusivity in filamentous fungal pel-
lets. Biotechnology and Bioengineering, 116(12): 3360–3371.
Kathrin Schrinner, Lukas Veiter, Stefan Schmideder, Philipp Doppler, Marcel Schrader,
Nadine Münch, Kristin Althof, Arno Kwade, Heiko Briesen, Christoph Herwig, and
Rainer Krull, 2020, Morphological and physiological characterization of filamentous
Lentzea aerocolonigenes: Comparison of biopellets by microscopy and flow cytome-
try. PLOS ONE, 15(6): e0234125.
Stefan Schmideder, Henri Müller, Lars Barthel, Tiaan Friedrich, Ludwig Niessen,
Vera Meyer, and Heiko Briesen, 2021, Universal law for diffusive mass transport through
mycelial networks. Biotechnology and Bioengineering, 118(2):930–943.
Oral Presentations
Stefan Schmideder, Christoph Kirse, Julia Hofinger, Sascha Rollié, and Heiko Briesen,
2018, Modeling the separation of microorganisms in bioprocesses by flotation. 6th
Population Balance Modelling Conference (PBM 2018), Gent, Belgium.
Kathrin Pommerehne, Marcel Schrader, Chrysoula Bliatsiou, Stefan Schmideder, Lutz
Böhm, Heiko Briesen, Matthias Kraume, Arno Kwade, and Rainer Krull1, 2018, Ex-
perimental and numerical investigations on cultivations of filamentous microorgan-
isms towards a better understanding and process control. ProcessNet Jahrestagung und
DECHEMA-Jahrestagung der Biotechnologen 2018, Aachen, Germany.
91
Heiko Briesen, Stefan Schmideder, Henri Müller, 2020, Morphological characteriza-
tion and modeling of filamentous fungi. 4th Indo-German Workshop on Advances in